VOLUMETRIC FEATURE FUSION BASED ON GEOMETRIC AND SIMILARITY CUES FOR THREE-DIMENSIONAL RECONSTRUCTION

Abstract

Disclosed are systems and techniques for three-dimensional reconstruction (3DR) of a scene. For example, a computing device can extract two-dimensional (2D) features from image frames of a scene. Each of the image frames includes a respective view of the scene. The computing device can unproject the 2D features from a 2D space onto a three-dimensional (3D) space to obtain 3D features. The computing device can determine weights for the 3D features based on geometric information and visual similarity information for the image frames. Each of the weights is associated with a respective feature 3D of the 3D features. The computing device can determine voxel features for the image frames based on the weights for the 3D features. The computing device can further form the 3DR of the scene based on the voxel features.

Description

FIELD

The present disclosure generally relates to image processing. For example, aspects of the present disclosure relate to enhanced volumetric feature fusion based on geometric and similarity cues for end-to-end three-dimensional reconstruction (3DR).

BACKGROUND

The increasing versatility of digital camera products has allowed digital cameras to be integrated into a wide array of devices and has expanded their use to different applications. For example, phones, drones, cars, computers, televisions, and many other devices today are often equipped with camera devices. The camera devices allow users to capture images and/or video (e.g., including frames of images) from any system equipped with a camera device. The images and/or videos can be captured for recreational use, professional photography, surveillance, and automation, among other applications. Moreover, camera devices are increasingly equipped with specific functionalities for modifying images or creating artistic effects on the images. For example, many camera devices are equipped with image processing capabilities for generating different effects on captured images.

In recent decades, there has been a demand for 3D content for computer graphics, virtual reality, and communications, triggering a change in emphasis for the requirements. Many existing systems for constructing 3D models are built around specialized hardware resulting in a high cost, which often cannot satisfy the requirements of these new applications. This need has stimulated the use of digital imaging facilities (e.g., cameras) for 3D reconstruction.

SUMMARY

The following presents a simplified summary relating to one or more aspects disclosed herein. Thus, the following summary should not be considered an extensive overview relating to all contemplated aspects, nor should the following summary be considered to identify key or critical elements relating to all contemplated aspects or to delineate the scope associated with any particular aspect. Accordingly, the following summary has the sole purpose to present certain concepts relating to one or more aspects relating to the mechanisms disclosed herein in a simplified form to precede the detailed description presented below.

Disclosed are systems, apparatuses, methods and computer-readable media for enhanced volumetric feature fusion based on geometric and similarity cues for end-to-end 3DR. According to at least one example, an apparatus for three-dimensional reconstruction (3DR) of a scene is provided. The apparatus includes at least one memory and at least one processor coupled to the at least one memory and configured to: extract a plurality of two-dimensional (2D) features from a plurality of image frames of a scene, wherein each image frame of the plurality of image frames comprises a respective view of the scene; unproject the plurality of 2D features from a 2D space onto a three-dimensional (3D) space to obtain a plurality of 3D features; determine a plurality of weights for the plurality of 3D features based on geometric information and visual similarity information for the plurality of image frames, wherein each weight of the plurality of weights is associated with a respective feature 3D of the plurality of 3D features; determine a plurality of voxel features for the plurality of image frames based on the plurality of weights for the plurality of 3D features; and generate the 3DR of the scene based on the plurality of voxel features.

In another illustrative example, a method is provided for three-dimensional reconstruction (3DR) of a scene. The method includes: extracting a plurality of two-dimensional (2D) features from a plurality of image frames of a scene, wherein each image frame of the plurality of image frames comprises a respective view of the scene; unprojecting the plurality of 2D features from a 2D space onto a three-dimensional (3D) space to obtain a plurality of 3D features; determining a plurality of weights for the plurality of 3D features based on geometric information and visual similarity information for the plurality of image frames, wherein each weight of the plurality of weights is associated with a respective feature 3D of the plurality of 3D features; determining a plurality of voxel features for the plurality of image frames based on the plurality of weights for the plurality of 3D features; and generating the 3DR of the scene based on the plurality of voxel features.

In another illustrative example, a non-transitory computer-readable medium is provided having stored thereon instructions that, when executed by one or more processors, cause the one or more processors to: extract a plurality of two-dimensional (2D) features from a plurality of image frames of a scene, wherein each image frame of the plurality of image frames comprises a respective view of the scene; unproject the plurality of 2D features from a 2D space onto a three-dimensional (3D) space to obtain a plurality of 3D features; determine a plurality of weights for the plurality of 3D features based on geometric information and visual similarity information for the plurality of image frames, wherein each weight of the plurality of weights is associated with a respective feature 3D of the plurality of 3D features; determine a plurality of voxel features for the plurality of image frames based on the plurality of weights for the plurality of 3D features; and generate the 3DR of the scene based on the plurality of voxel features.

In another illustrative example, an apparatus for three-dimensional reconstruction (3DR) of a scene is provided. The apparatus includes: means for extracting a plurality of two-dimensional (2D) features from a plurality of image frames of a scene, wherein each image frame of the plurality of image frames comprises a respective view of the scene; means for unprojecting the plurality of 2D features from a 2D space onto a three-dimensional (3D) space to obtain a plurality of 3D features; means for determining a plurality of weights for the plurality of 3D features based on geometric information and visual similarity information for the plurality of image frames, wherein each weight of the plurality of weights is associated with a respective feature 3D of the plurality of 3D features; means for determining a plurality of voxel features for the plurality of image frames based on the plurality of weights for the plurality of 3D features; and means for generating the 3DR of the scene based on the plurality of voxel features.

Aspects generally include a method, apparatus, system, computer program product, non-transitory computer-readable medium, user device, image processing device, and/or processing system as substantially described with reference to and as illustrated by the drawings and specification.

Some aspects include a device having a processor configured to perform one or more operations of any of the methods summarized above. Further aspects include processing devices for use in a device configured with processor-executable instructions to perform operations of any of the methods summarized above. Further aspects include a non-transitory processor-readable storage medium having stored thereon processor-executable instructions configured to cause a processor of a device to perform operations of any of the methods summarized above. Further aspects include a device having means for performing functions of any of the methods summarized above.

In some aspects, one or more of the apparatuses described herein is, can be part of, or can include a mobile device, a smart or connected device, a camera system, and/or an extended reality (XR) device (e.g., a virtual reality (VR) device, an augmented reality (AR) device, or a mixed reality (MR) device). In some examples, the apparatus(es) can include or be part of a vehicle, a mobile device (e.g., a mobile telephone or so-called “smart phone” or other mobile device), a wearable device, a personal computer, a laptop computer, a tablet computer, a server computer, a robotics device or system, an aviation system, or other device. In some aspects, the apparatus(es) include an image sensor (e.g., a camera) or multiple image sensors (e.g., multiple cameras) for capturing one or more images. In some aspects, the apparatus(es) include one or more displays for displaying one or more images, notifications, and/or other displayable data. In some aspects, the apparatus(es) include one or more speakers, one or more light-emitting devices, and/or one or more microphones. In some aspects, the apparatuses(es) can include one or more sensors. In some cases, the one or more sensors can be used for determining a location of the apparatuses, a state of the apparatuses (e.g., a tracking state, an operating state, a temperature, a humidity level, and/or other state), and/or for other purposes.

Some aspects include a device having a processor configured to perform one or more operations of any of the methods summarized above. Further aspects include processing devices for use in a device configured with processor-executable instructions to perform operations of any of the methods summarized above. Further aspects include a non-transitory processor-readable storage medium having stored thereon processor-executable instructions configured to cause a processor of a device to perform operations of any of the methods summarized above. Further aspects include a device having means for performing functions of any of the methods summarized above.

The foregoing has outlined rather broadly the features and technical advantages of examples according to the disclosure in order that the detailed description that follows may be better understood. Additional features and advantages will be described hereinafter. The conception and specific examples disclosed may be readily utilized as a basis for modifying or designing other structures for carrying out the same purposes of the present disclosure. Such equivalent constructions do not depart from the scope of the appended claims. Characteristics of the concepts disclosed herein, both their organization and method of operation, together with associated advantages will be better understood from the following description when considered in connection with the accompanying figures. Each of the figures is provided for the purposes of illustration and description, and not as a definition of the limits of the claims. The foregoing, together with other features and aspects, will become more apparent upon referring to the following specification, claims, and accompanying drawings.

This summary is not intended to identify key or essential features of the claimed subject matter, nor is it intended to be used in isolation to determine the scope of the claimed subject matter. The subject matter should be understood by reference to appropriate portions of the entire specification of this patent, any or all drawings, and each claim.

The preceding, together with other features and embodiments, will become more apparent upon referring to the following specification, claims, and accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

Illustrative aspects of the present application are described in detail below with reference to the following figures:

FIG. 1 is a block diagram illustrating an example architecture of an image capture and processing system, in accordance with some aspects of the present disclosure.

FIG. 2 is a block diagram illustrating an example of interactions between components of an image capture and processing system, in accordance with some aspects of the present disclosure.

FIG. 3 is a block diagram illustrating an example device that may employ a color metadata buffer for 3DR, in accordance with some aspects of the present disclosure.

FIG. 4 is a diagram illustrating an example of a 3D surface reconstruction of a scene modeled as a volume grid, in accordance with some aspects of the present disclosure.

FIG. 5 is a diagram illustrating an example of a hash mapping function for indexing blocks (e.g., voxels) in a volume grid, in accordance with some aspects of the present disclosure.

FIG. 6 is a diagram illustrating an example of a block (e.g., a voxel), in accordance with some aspects of the present disclosure.

FIG. 7 is a diagram illustrating an example of a truncated signed distance function (TSDF) volume reconstruction, in accordance with some aspects of the present disclosure.

FIG. 8 is a diagram illustrating an example of a 3DR pipeline, in accordance with some aspects of the present disclosure.

FIG. 9 is a diagram illustrating an example of a verification network that may be employed for 3DR, in accordance with some aspects of the present disclosure.

FIG. 10 is a table illustrating examples of 3DR performance results using a simple averaging of features versus using a weighted average of features for the 3DR, in accordance with some aspects of the present disclosure.

FIG. 11 is a diagram illustrating examples of visual 3DR results using a simple averaging of features versus using a weighted average of features for the 3DR, in accordance with some aspects of the present disclosure.

FIG. 12 is a flow diagram illustrating an example of a process for enhanced volumetric feature fusion based on geometric and similarity cues for end-to-end 3DR, in accordance with some aspects of the present disclosure.

FIG. 13 is a diagram illustrating an example of a system for implementing certain aspects described herein, in accordance with some aspects of the present disclosure.

DETAILED DESCRIPTION

Certain aspects of this disclosure are provided below for illustration purposes. Alternate aspects may be devised without departing from the scope of the disclosure. Additionally, well-known elements of the disclosure will not be described in detail or will be omitted so as not to obscure the relevant details of the disclosure. Some of the aspects described herein can be applied independently and some of them may be applied in combination as would be apparent to those of skill in the art. In the following description, for the purposes of explanation, specific details are set forth in order to provide a thorough understanding of aspects of the application. However, it will be apparent that various aspects may be practiced without these specific details. The figures and description are not intended to be restrictive.

The ensuing description provides example aspects only, and is not intended to limit the scope, applicability, or configuration of the disclosure. Rather, the ensuing description of the example aspects will provide those skilled in the art with an enabling description for implementing an example aspect. It should be understood that various changes may be made in the function and arrangement of elements without departing from the spirit and scope of the application as set forth in the appended claims.

The terms “exemplary” and/or “example” are used herein to mean “serving as an example, instance, or illustration.” Any aspect described herein as “exemplary” and/or “example” is not necessarily to be construed as preferred or advantageous over other aspects. Likewise, the term “aspects of the disclosure” does not require that all aspects of the disclosure include the discussed feature, advantage or mode of operation.

A camera is a device that receives light and captures image frames, such as still images or video frames, using an image sensor. The terms “image,” “image frame,” and “frame” are used interchangeably herein. Cameras may include processors, such as image signal processors (ISPs), that can receive one or more image frames and process the one or more image frames. For example, a raw image frame captured by a camera sensor can be processed by an ISP to generate a final image. Processing by the ISP can be performed by a plurality of filters or processing blocks being applied to the captured image frame, such as denoising or noise filtering, edge enhancement, color balancing, contrast, intensity adjustment (such as darkening or lightening), tone adjustment, among others. Image processing blocks or modules may include lens/sensor noise correction, Bayer filters, de-mosaicing, color conversion, correction or enhancement/suppression of image attributes, denoising filters, sharpening filters, among others.

Cameras can be configured with a variety of image capture and image processing operations and settings. The different settings result in images with different appearances. Some camera operations are determined and applied before or during capture of the image, such as automatic exposure control (AEC) and automatic white balance (AWB) processing. Additional camera operations applied before, during, or after capture of an image include operations involving zoom (e.g., zooming in or out), ISO, aperture size, f/stop, shutter speed, and gain. Other camera operations can configure post-processing of an image, such as alterations to contrast, brightness, saturation, sharpness, levels, curves, or colors.

As previously mentioned, in recent decades, there has been a demand for three-dimensional (3D) content for computer graphics, virtual reality, and communications, triggering a change in emphasis for the requirements. Many existing systems for constructing 3D models are built around specialized hardware resulting in a high cost, and often cannot satisfy the requirements of these new applications. The requirements have stimulated the use of digital imaging (e.g., using images from cameras) for 3D reconstruction.

In some cases, volume blocks (e.g., voxels) can be utilized to reconstruct a 3D scene from two-dimensional (2D) images, such as stereo images obtained from a stereo camera. A voxel represents a value on a regular grid in 3D space. As with pixels in a 2D bitmap, voxels do not have their position (e.g., coordinates) explicitly encoded within their values. Instead, rendering systems infer the position of a voxel based upon its position relative to other voxels (e.g., its position in the data structure that makes up a single volumetric image).

In some examples, a system can perform 3D reconstruction (3DR) using depth frames and an associated live camera pose estimate for 3D scene reconstruction. In 3D surface reconstruction, the system can model the scene as a 3D sparse volumetric representation (e.g., referred to as a volume grid). The volume grid can contain a set of voxel blocks, which are each indexed by their position in space with a sparse data representation (e.g., only storing blocks that surround an object and/or obstacle).

In one illustrative example, a system can perform 3DR to reconstruct a 3D scene from 2D depth frames and color frames. The system can divide the scene into 3D blocks (e.g., voxels or voxel blocks, as noted previously). For example, the system may project each voxel onto a 2D depth frame and a 2D color frame to determine the depth and color of the voxel. Once all of the voxels that refer to (e.g., are associated with) this depth frame and color frame are updated accordingly, the process can repeat for a new depth frame and color frame pair or set.

In 3DR, 3D scenes are represented using a 3D volume of points called voxels, where each voxel typically carries implicit surface information, such as in the form of a truncated Signed Distance Function (TSDF) value and a weight for depth integration. The TSDF value is a measure of distance of the voxel from a surface, and the weight is a measure of the reliability of the TSDF value. A TSDF weight can be estimated using various approaches, such as a simple counter (e.g., a binary weight of 1 or 0), based on a depth range, or from a confidence of the depth predictions.

Currently, end-to-end multi-view 3D reconstruction using deep learning has recently gained significant attention due to promising results shown by recent works. The main motivation of these works is to compute 2D visual features, which are then lifted into a 3D volume of voxels. However, as there are multiple views of a scene, typically, features are simply averaged to get to a single embedding, which can result in over-smoothed surfaces in the 3DR. Follow-up works have addressed this issue of over-smoothed surfaces by introducing expensive transformer layers to compute cross-attention amongst view features. Follow-up works have also addressed this issue by incurring additional costs to first estimate the depth, which is then used to determine how the features should be aggregated. As such, systems and techniques for improved volumetric feature fusion for end-to-end 3DR can be beneficial.

In one or more aspects, systems, apparatuses, processes (also referred to as methods), and computer-readable media (collectively referred to herein as “systems and techniques”) are described herein for providing enhanced volumetric feature fusion based on geometric and similarity cues for end-to-end 3DR. For instance, in one or more examples, the systems and techniques use available geometric cues (e.g., the distance of voxels from camera centers, and angles between rays going through consecutive camera centers from a voxel) as well as a similarity of view features to learn adaptive scalars to weight corresponding view features. In some cases, the systems and techniques can ingest these cues and output appropriate weights (e.g., normalized weights) that can better aggregate the features to result in a higher 3D reconstruction quality, without introducing additional components with requiring significant computation. In one illustrative example, the systems and techniques can use a multi-layer perceptron (MLP) network, a fully connected layer, or other network component or layer to ingest the cues and generate the weights.

In one or more examples, the systems and techniques employ an MLP, a fully connected layer, or other component or layer that is added or trained with 2D feature similarities across views and geometric cues (e.g., the distances of voxels from camera centers, and angles between rays going through consecutive camera centers from a voxel) as input, and weights (e.g., normalized weights, which may be scalar values) for the respective features as output. The final 3D reconstruction is based on the weighted average (using the weights) of the voxel features that are projected inside the camera plane. In some examples, the systems and techniques can also utilize additional information, such as semantic labels, depth prediction, depth uncertainties, and 3D scans to enhance the multi-view feature aggregation. The systems and techniques can also use neighboring voxel's feature similarity across views to improve the robustness of the weighing mechanism. In some examples, the systems and techniques can employ a lookup table (LUT) to map the geometric cues to weights after training the MLP to reduce the computation complexity.

In some examples, the systems and techniques can apply to XR use cases. For XR use cases, a head-mounted device (HMD), for example, with a limited power and computation budget can better aggregate visual features across multiple samples from a scene via incorporation of geometric and visual similarity cues and, hence, be able to reconstruct 3D surfaces with more accuracy.

In one or more aspects, the systems and techniques incorporate inherent geometric information that is available in multi-view 3D scene reconstruction without the need of any additional dedicated network and complex computing modules. In the existing mainstream solutions, features computed from multiple views of a scene for a given voxel are simply averaged, resulting in an over-smoothed reconstruction. For example, assuming there are multiple views of a scene, for a given voxel, a lifted feature (e.g., in existing mainstream solutions) is obtained via a simple averaging across views. The systems and techniques replace this simple averaging scheme by utilizing an adaptive weighted average mechanism that can determine how much each view feature should contribute, depending upon geometric cues and visual similarity of computed features.

The systems and techniques have several technical advantages. For example, the systems and techniques leverage natural geometric cues that are readily available, without the need for any additional task-specific modules, such has for depth estimation. For another example, the systems and techniques utilize distance, view angles, and feature similarity to allow for the network to better infer the geometry of a scene. For another example, the systems and techniques can improve the accuracy of reconstructed surfaces by using only multi-view red, green, blue (RGB) information. The systems and techniques do not require complex operations that may be challenging to implement on a target mobile device (e.g., employing a camera).

In one or more examples, the systems and techniques provide a method for 3DR of a scene. During operation of the method, a plurality of two-dimensional (2D) features can be extracted from a plurality of image frames of a scene of an environment. Each image frame of the plurality of image frames can include a respective view of the scene. The 2D features are unprojected to a three-dimensional 3D space by assigning each 2D feature to a 3D location (referred to as a voxel). There will be multiple unprojected features (referred to as voxel features) for each individual voxel (or 3D location) since there are multiple frames, and the multiple voxel features for each voxel can be aggregated into one voxel feature. In contrast to performing the averaging technique described previously, the systems and techniques can calculate a plurality of weights (e.g., normalized weights) adaptively for the multiple features to perform weighted averaging of the voxel features to obtain a voxel feature for each voxel. The resulting voxel features for the voxels of the 3D space are referred to as a feature volume. Once the voxel features are obtained, the 3D reconstruction can be performed on top of the feature volume.

In some cases, the plurality of weights (e.g., the normalized weights) for the multiple voxel features can be determined based on geometric information and visual similarity information for the plurality of image frames. Each weight of the plurality of weights can be associated with a respective feature of the multiple features (the features for a given 3D location or voxel). As noted above, a plurality of voxel features for the plurality of image frames can be determined based on the plurality of weights for the plurality of features.

In some examples, the visual similarity information can be based on a visual similarity of two or more features of the plurality of features from different images frames of the plurality of image frames. The geometric information may be based on a plurality of distances. Each distance of the plurality of distances can be between a respective a voxel (corresponding to a 3D location in a scene) of a plurality of voxels to a location of a respective image sensor (e.g., camera) used to capture the image frames. In one or more examples, the respective image sensor may be a camera. In some examples, the geometric information may be based on a plurality of view angles. Each view angle of the plurality of view angles can be between two respective rays each associated with a respective image frame of the plurality of image frames.

In one or more examples, determining the plurality of weights for the plurality of features may be further based on semantic labels, depth predictions, depth uncertainties, and/or 3D scans associated with the plurality of features. In some examples, the method can further involve grouping one or more voxel features of the plurality of voxel features together to form a group of voxel features (e.g., a patch of voxel features), based on the one or more voxel features having a visual similarity across different views of the scene. The unprojecting of the plurality of features from the 2D space onto the 3D space can then be based on the group of voxel features (e.g., the patch of voxel features). In one or more examples, determining the plurality of weights for the plurality of features may be further based on a lookup table (LUT) mapping the geometric information to weights of the plurality of weights. In one or more examples, each weight of the plurality of weights is a positive value. In some examples, a sum of all of the weights of the plurality of weights is equal to one.

Additional aspects of the present disclosure are described in more detail below.

FIG. 1 is a block diagram illustrating an architecture of an image capture and processing system 100. The image capture and processing system 100 includes various components that are used to capture and process images of scenes (e.g., an image of a scene 110). The image capture and processing system 100 can capture standalone images (or photographs) and/or can capture videos that include multiple images (or video frames) in a particular sequence. A lens 115 of the system 100 faces a scene 110 and receives light from the scene 110. The lens 115 bends the light toward the image sensor 130. The light received by the lens 115 passes through an aperture controlled by one or more control mechanisms 120 and is received by an image sensor 130.

The one or more control mechanisms 120 may control exposure, focus, and/or zoom based on information from the image sensor 130 and/or based on information from the image processor 150. The one or more control mechanisms 120 may include multiple mechanisms and components; for instance, the control mechanisms 120 may include one or more exposure control mechanisms 125A, one or more focus control mechanisms 125B, and/or one or more zoom control mechanisms 125C. The one or more control mechanisms 120 may also include additional control mechanisms besides those that are illustrated, such as control mechanisms controlling analog gain, flash, HDR, depth of field, and/or other image capture properties.

The focus control mechanism 125B of the control mechanisms 120 can obtain a focus setting. In some examples, focus control mechanism 125B store the focus setting in a memory register. Based on the focus setting, the focus control mechanism 125B can adjust the position of the lens 115 relative to the position of the image sensor 130. For example, based on the focus setting, the focus control mechanism 125B can move the lens 115 closer to the image sensor 130 or farther from the image sensor 130 by actuating a motor or servo, thereby adjusting focus. In some cases, additional lenses may be included in the device 105A, such as one or more microlenses over each photodiode of the image sensor 130, which each bend the light received from the lens 115 toward the corresponding photodiode before the light reaches the photodiode. The focus setting may be determined via contrast detection autofocus (CDAF), phase detection autofocus (PDAF), or some combination thereof. The focus setting may be determined using the control mechanism 120, the image sensor 130, and/or the image processor 150. The focus setting may be referred to as an image capture setting and/or an image processing setting.

The exposure control mechanism 125A of the control mechanisms 120 can obtain an exposure setting. In some cases, the exposure control mechanism 125A stores the exposure setting in a memory register. Based on this exposure setting, the exposure control mechanism 125A can control a size of the aperture (e.g., aperture size or f/stop), a duration of time for which the aperture is open (e.g., exposure time or shutter speed), a sensitivity of the image sensor 130 (e.g., ISO speed or film speed), analog gain applied by the image sensor 130, or any combination thereof. The exposure setting may be referred to as an image capture setting and/or an image processing setting.

The zoom control mechanism 125C of the control mechanisms 120 can obtain a zoom setting. In some examples, the zoom control mechanism 125C stores the zoom setting in a memory register. Based on the zoom setting, the zoom control mechanism 125C can control a focal length of an assembly of lens elements (lens assembly) that includes the lens 115 and one or more additional lenses. For example, the zoom control mechanism 125C can control the focal length of the lens assembly by actuating one or more motors or servos to move one or more of the lenses relative to one another. The zoom setting may be referred to as an image capture setting and/or an image processing setting. In some examples, the lens assembly may include a parfocal zoom lens or a varifocal zoom lens. In some examples, the lens assembly may include a focusing lens (which can be lens 115 in some cases) that receives the light from the scene 110 first, with the light then passing through an afocal zoom system between the focusing lens (e.g., lens 115) and the image sensor 130 before the light reaches the image sensor 130. The afocal zoom system may, in some cases, include two positive (e.g., converging, convex) lenses of equal or similar focal length (e.g., within a threshold difference) with a negative (e.g., diverging, concave) lens between them. In some cases, the zoom control mechanism 125C moves one or more of the lenses in the afocal zoom system, such as the negative lens and one or both of the positive lenses.

The image sensor 130 includes one or more arrays of photodiodes or other photosensitive elements. Each photodiode measures an amount of light that eventually corresponds to a particular pixel in the image produced by the image sensor 130. In some cases, different photodiodes may be covered by different color filters, and may thus measure light matching the color of the filter covering the photodiode. For instance, Bayer color filters include red color filters, blue color filters, and green color filters, with each pixel of the image generated based on red light data from at least one photodiode covered in a red color filter, blue light data from at least one photodiode covered in a blue color filter, and green light data from at least one photodiode covered in a green color filter. Other types of color filters may use yellow, magenta, and/or cyan (also referred to as “emerald”) color filters instead of or in addition to red, blue, and/or green color filters. Some image sensors may lack color filters altogether, and may instead use different photodiodes throughout the pixel array (in some cases vertically stacked). The different photodiodes throughout the pixel array can have different spectral sensitivity curves, therefore responding to different wavelengths of light. Monochrome image sensors may also lack color filters and therefore lack color depth.

In some cases, the image sensor 130 may alternately or additionally include opaque and/or reflective masks that block light from reaching certain photodiodes, or portions of certain photodiodes, at certain times and/or from certain angles, which may be used for phase detection autofocus (PDAF). The image sensor 130 may also include an analog gain amplifier to amplify the analog signals output by the photodiodes and/or an analog to digital converter (ADC) to convert the analog signals output of the photodiodes (and/or amplified by the analog gain amplifier) into digital signals. In some cases, certain components or functions discussed with respect to one or more of the control mechanisms 120 may be included instead or additionally in the image sensor 130. The image sensor 130 may be a charge-coupled device (CCD) sensor, an electron-multiplying CCD (EMCCD) sensor, an active-pixel sensor (APS), a complimentary metal-oxide semiconductor (CMOS), an N-type metal-oxide semiconductor (NMOS), a hybrid CCD/CMOS sensor (e.g., sCMOS), or some other combination thereof.

The image processor 150 may include one or more processors, such as one or more image signal processors (ISPs) (including ISP 154), one or more host processors (including host processor 152), and/or one or more of any other type of processor 1210 discussed with respect to the computing system 1200. The host processor 152 can be a digital signal processor (DSP) and/or other type of processor. In some implementations, the image processor 150 is a single integrated circuit or chip (e.g., referred to as a system-on-chip or SoC) that includes the host processor 152 and the ISP 154. In some cases, the chip can also include one or more input/output ports (e.g., input/output (I/O) ports 156), central processing units (CPUs), graphics processing units (GPUs), broadband modems (e.g., 3G, 4G or LTE, 5G, etc.), memory, connectivity components (e.g., Bluetooth™, Global Positioning System (GPS), etc.), any combination thereof, and/or other components. The I/O ports 156 can include any suitable input/output ports or interface according to one or more protocol or specification, such as an Inter-Integrated Circuit 2 (I2C) interface, an Inter-Integrated Circuit 3 (I3C) interface, a Serial Peripheral Interface (SPI) interface, a serial General Purpose Input/Output (GPIO) interface, a Mobile Industry Processor Interface (MIPI) (such as a MIPI CSI-2 physical (PHY) layer port or interface, an Advanced High-performance Bus (AHB) bus, any combination thereof, and/or other input/output port. In one illustrative example, the host processor 152 can communicate with the image sensor 130 using an I2C port, and the ISP 154 can communicate with the image sensor 130 using an MIPI port.

The image processor 150 may perform a number of tasks, such as de-mosaicing, color space conversion, image frame downsampling, pixel interpolation, automatic exposure (AE) control, automatic gain control (AGC), CDAF, PDAF, automatic white balance, merging of image frames to form an HDR image, image recognition, object recognition, feature recognition, receipt of inputs, managing outputs, managing memory, or some combination thereof. The image processor 150 may store image frames and/or processed images in random access memory (RAM) 140/1120, read-only memory (ROM) 145/1125, a cache 1112, a memory unit 1115, another storage device 1130, or some combination thereof.

Various input/output (I/O) devices 160 may be connected to the image processor 150. The I/O devices 160 can include a display screen, a keyboard, a keypad, a touchscreen, a trackpad, a touch-sensitive surface, a printer, any other output devices 1135, any other input devices 1145, or some combination thereof. In some cases, a caption may be input into the image processing device 105B through a physical keyboard or keypad of the I/O devices 160, or through a virtual keyboard or keypad of a touchscreen of the I/O devices 160. The I/O 160 may include one or more ports, jacks, or other connectors that enable a wired connection between the device 105B and one or more peripheral devices, over which the device 105B may receive data from the one or more peripheral device and/or transmit data to the one or more peripheral devices. The I/O 160 may include one or more wireless transceivers that enable a wireless connection between the device 105B and one or more peripheral devices, over which the device 105B may receive data from the one or more peripheral device and/or transmit data to the one or more peripheral devices. The peripheral devices may include any of the previously-discussed types of I/O devices 160 and may themselves be considered I/O devices 160 once they are coupled to the ports, jacks, wireless transceivers, or other wired and/or wireless connectors.

In some cases, the image capture and processing system 100 may be a single device. In some cases, the image capture and processing system 100 may be two or more separate devices, including an image capture device 105A (e.g., a camera) and an image processing device 105B (e.g., a computing device coupled to the camera). In some implementations, the image capture device 105A and the image processing device 105B may be coupled together, for example via one or more wires, cables, or other electrical connectors, and/or wirelessly via one or more wireless transceivers. In some implementations, the image capture device 105A and the image processing device 105B may be disconnected from one another.

As shown in FIG. 1, a vertical dashed line divides the image capture and processing system 100 of FIG. 1 into two portions that represent the image capture device 105A and the image processing device 105B, respectively. The image capture device 105A includes the lens 115, control mechanisms 120, and the image sensor 130. The image processing device 105B includes the image processor 150 (including the ISP 154 and the host processor 152), the RAM 140, the ROM 145, and the I/O 160. In some cases, certain components illustrated in the image capture device 105A, such as the ISP 154 and/or the host processor 152, may be included in the image capture device 105A.

The image capture and processing system 100 can include an electronic device, such as a mobile or stationary telephone handset (e.g., smartphone, cellular telephone, or the like), a desktop computer, a laptop or notebook computer, a tablet computer, a set-top box, a television, a camera, a display device, a digital media player, a video gaming console, a video streaming device, an Internet Protocol (IP) camera, or any other suitable electronic device. In some examples, the image capture and processing system 100 can include one or more wireless transceivers for wireless communications, such as cellular network communications, 802.11 wi-fi communications, wireless local area network (WLAN) communications, or some combination thereof. In some implementations, the image capture device 105A and the image processing device 105B can be different devices. For instance, the image capture device 105A can include a camera device and the image processing device 105B can include a computing device, such as a mobile handset, a desktop computer, or other computing device.

While the image capture and processing system 100 is shown to include certain components, one of ordinary skill will appreciate that the image capture and processing system 100 can include more components than those shown in FIG. 1. The components of the image capture and processing system 100 can include software, hardware, or one or more combinations of software and hardware. For example, in some implementations, the components of the image capture and processing system 100 can include and/or can be implemented using electronic circuits or other electronic hardware, which can include one or more programmable electronic circuits (e.g., microprocessors, GPUs, DSPs, CPUs, and/or other suitable electronic circuits), and/or can include and/or be implemented using computer software, firmware, or any combination thereof, to perform the various operations described herein. The software and/or firmware can include one or more instructions stored on a computer-readable storage medium and executable by one or more processors of the electronic device implementing the image capture and processing system 100.

The host processor 152 can configure the image sensor 130 with new parameter settings (e.g., via an external control interface such as I2C, I3C, SPI, GPIO, and/or other interface). In one illustrative example, the host processor 152 can update exposure settings used by the image sensor 130 based on internal processing results of an exposure control algorithm from past image frames.

In some examples, the host processor 152 can perform electronic image stabilization (EIS). For instance, the host processor 152 can determine a motion vector corresponding to motion compensation for one or more image frames. In some aspects, host processor 152 can position a cropped pixel array (“the image window”) within the total array of pixels. The image window can include the pixels that are used to capture images. In some examples, the image window can include all of the pixels in the sensor, except for a portion of the rows and columns at the periphery of the sensor. In some cases, the image window can be in the center of the sensor while the image capture device 105A is stationary. In some aspects, the peripheral pixels can surround the pixels of the image window and form a set of buffer pixel rows and buffer pixel columns around the image window. Host processor 152 can implement EIS and shift the image window from frame to frame of video, so that the image window tracks the same scene over successive frames (e.g., assuming that the subject does not move). In some examples in which the subject moves, host processor 152 can determine that the scene has changed.

In some examples, the image window can include at least 95% (e.g., 95% to 99%) of the pixels on the sensor. The first region of interest (ROI) (e.g., used for AE and/or AWB) may include the image data within the field of view of at least 95% (e.g., 95% to 99%) of the plurality of imaging pixels in the image sensor 130 of the image capture device 105A. In some aspects, a number of buffer pixels at the periphery of the sensor (outside of the image window) can be reserved as a buffer to allow the image window to shift to compensate for jitter. In some cases, the image window can be moved so that the subject remains at the same location within the adjusted image window, even though light from the subject may impinge on a different region of the sensor. In another example, the buffer pixels can include the ten topmost rows, ten bottommost rows, ten leftmost columns and ten rightmost columns of pixels on the sensor. In some configurations, the buffer pixels are not used for AF, AE or AWB when the image capture device 105A is stationary and the buffer pixels not included in the image output. If jitter moves the sensor to the left by twice the width of a column of pixels between frames, the EIS algorithm can be used to shift the image window to the right by two columns of pixels, so the captured image shows the same scene in the next frame as in the current frame. Host processor 152 can use EIS to smoothen the transition from one frame to the next.

In some aspects, the host processor 152 can also dynamically configure the parameter settings of the internal pipelines or modules of the ISP 154 to match the settings of one or more input image frames from the image sensor 130 so that the image data is correctly processed by the ISP 154. Processing (or pipeline) blocks or modules of the ISP 154 can include modules for lens/sensor noise correction, de-mosaicing, color conversion, correction or enhancement/suppression of image attributes, denoising filters, sharpening filters, among others. The settings of different modules of the ISP 154 can be configured by the host processor 152. Each module may include a large number of tunable parameter settings. Additionally, modules may be co-dependent as different modules may affect similar aspects of an image. For example, denoising and texture correction or enhancement may both affect high frequency aspects of an image. As a result, a large number of parameters are used by an ISP to generate a final image from a captured raw image.

In some cases, the image capture and processing system 100 may perform one or more of the image processing functionalities described above automatically. For instance, one or more of the control mechanisms 120 may be configured to perform auto-focus operations, auto-exposure operations, and/or auto-white-balance operations. In some embodiments, an auto-focus functionality allows the image capture device 105A to focus automatically prior to capturing the desired image. Various auto-focus technologies exist. For instance, active autofocus technologies determine a range between a camera and a subject of the image via a range sensor of the camera, typically by emitting infrared lasers or ultrasound signals and receiving reflections of those signals. In addition, passive auto-focus technologies use a camera's own image sensor to focus the camera, and thus do not require additional sensors to be integrated into the camera. Passive AF techniques include Contrast Detection Auto Focus (CDAF), Phase Detection Auto Focus (PDAF), and in some cases hybrid systems that use both. The image capture and processing system 100 may be equipped with these or any additional type of auto-focus technology.

Synchronization between the image sensor 130 and the ISP 154 is important in order to provide an operational image capture system that generates high quality images without interruption and/or failure. FIG. 2 is a block diagram illustrating an example of an image capture and processing system 200 including an image processor 250 (including host processor 252 and ISP 254) in communication with an image sensor 230. The configuration shown in FIG. 2 is illustrative of traditional synchronization techniques used in camera systems. In general, the host processor 252 attempts to provide synchronization between the image sensor 230 and the ISP 254 using fixed periods of time by separately communicating with the image sensor 230 and the ISP 254. For example, in traditional camera systems, the host processor 252 communicates with the image sensor 230 (e.g., over an I2C port) and programs the image sensor 230 parameters with a first fixed period of time, such as 2-frame periods ahead of when that image frame will be processed by the ISP 254. The host processor 252 communicates with the ISP 254 (e.g., over an internal AHB bus or other interface) and programs the ISP 254 parameter settings with a second fixed period of time, such as 1-frame period ahead of when that image frame will be processed by the ISP 254.

The image sensor 230 can send image frames to the ISP 254 (B-to-C in FIG. 2), such as over an MIPI CSI-2 PHY port or interface, or other suitable interface. However, the communication between the host processor 252 and the image sensor 230 (shown as from A to B) is undeterministic. Similarly, the communication between the image sensor 230 and the ISP 254 (shown as from B to C) and the communication the host processor 252 and the ISP 254 (shown as from A to C) are also undeterministic. For example, there can be varying latencies in programming of the image sensor 230 and the ISP 254 by the host processor 252, which can result in a parameter settings mismatch between the sensor and the ISP. The latencies can be due to high CPU usage, congestion in one or more I/O ports, and/or due to other factors.

FIG. 3 is a block diagram of an example device 300 that may employ a color metadata buffer for 3D reconstruction. Device 300 may include or may be coupled to a camera 302, and may further include a processor 306, a memory 308 storing instructions 310, a camera controller 312, a display 316, and a number of input/output (I/O) components 318 including one or more microphones (not shown). The example device 300 may be any suitable device capable of capturing and/or storing images or video including, for example, wired and wireless communication devices (such as camera phones, smartphones, tablets, security systems, smart home devices, connected home devices, surveillance devices, internet protocol (IP) devices, dash cameras, laptop computers, desktop computers, automobiles, drones, aircraft, and so on), digital cameras (including still cameras, video cameras, and so on), or any other suitable device. The device 300 may include additional features or components not shown. For example, a wireless interface, which may include a number of transceivers and a baseband processor, May be included for a wireless communication device. Device 300 may include or may be coupled to additional cameras other than the camera 302. The disclosure should not be limited to any specific examples or illustrations, including the example device 300.

Camera 302 may be capable of capturing individual image frames (such as still images) and/or capturing video (such as a succession of captured image frames). Camera 302 may include one or more image sensors (not shown for simplicity) and shutters for capturing an image frame and providing the captured image frame to camera controller 312. Although a single camera 302 is shown, any number of cameras or camera components may be included and/or coupled to device 300. For example, the number of cameras may be increased to achieve greater depth determining capabilities or better resolution for a given FOV.

Memory 308 may be a non-transient or non-transitory computer readable medium storing computer-executable instructions 310 to perform all or a portion of one or more operations described in this disclosure. Device 300 may also include a power supply 320, which may be coupled to or integrated into the device 300.

Processor 306 may be one or more suitable processors capable of executing scripts or instructions of one or more software programs (such as the instructions 310) stored within memory 308. In some aspects, processor 306 may be one or more general purpose processors that execute instructions 310 to cause device 300 to perform any number of functions or operations. In additional or alternative aspects, processor 306 may include integrated circuits or other hardware to perform functions or operations without the use of software. While shown to be coupled to each other via processor 306 in the example of FIG. 3, processor 306, memory 308, camera controller 312, display 316, and I/O components 318 may be coupled to one another in various arrangements. For example, processor 306, memory 308, camera controller 312, display 316, and/or I/O components 318 may be coupled to each other via one or more local buses (not shown for simplicity).

Display 316 may be any suitable display or screen allowing for user interaction and/or to present items (such as captured images and/or videos) for viewing by the user. In some aspects, display 316 may be a touch-sensitive display. Display 316 may be part of or external to device 300. Display 316 may comprise an LCD, LED, OLED, or similar display. I/O components 318 may be or may include any suitable mechanism or interface to receive input (such as commands) from the user and/or to provide output to the user. For example, I/O components 318 may include (but are not limited to) a graphical user interface, keyboard, mouse, microphone and speakers, and so on.

Camera controller 312 may include an image signal processor (ISP) 314, which may be (or may include) one or more image signal processors to process captured image frames or videos provided by camera 302. For example, ISP 314 may be configured to perform various processing operations for automatic focus (AF), automatic white balance (AWB), and/or automatic exposure (AE), which may also be referred to as automatic exposure control (AEC). Examples of image processing operations include, but are not limited to, cropping, scaling (e.g., to a different resolution), image stitching, image format conversion, color interpolation, image interpolation, color processing, image filtering (e.g., spatial image filtering), and/or the like.

In some example implementations, camera controller 312 (such as the ISP 314) may implement various functionality, including imaging processing and/or control operation of camera 302. In some aspects, ISP 314 may execute instructions from a memory (such as instructions 310 stored in memory 308 or instructions stored in a separate memory coupled to ISP 314) to control image processing and/or operation of camera 302. In other aspects, ISP 314 may include specific hardware to control image processing and/or operation of camera 302. ISP 314 may alternatively or additionally include a combination of specific hardware and the ability to execute software instructions.

While not shown in FIG. 3, in some implementations, ISP 314 and/or camera controller 312 may include an AF module, an AWB module, and/or an AE module. ISP 314 and/or camera controller 312 may be configured to execute an AF process, an AWB process, and/or an AE process. In some examples, ISP 314 and/or camera controller 312 may include hardware-specific circuits (e.g., an application-specific integrated circuit (ASIC)) configured to perform the AF, AWB, and/or AE processes. In other examples, ISP 314 and/or camera controller 312 may be configured to execute software and/or firmware to perform the AF, AWB, and/or AE processes. When configured in software, code for the AF, AWB, and/or AE processes may be stored in memory (such as instructions 310 stored in memory 308 or instructions stored in a separate memory coupled to ISP 314 and/or camera controller 312). In other examples, ISP 314 and/or camera controller 312 may perform the AF, AWB, and/or AE processes using a combination of hardware, firmware, and/or software. When configured as software, AF, AWB, and/or AE processes may include instructions that configure ISP 314 and/or camera controller 312 to perform various image processing and device managements tasks, including the techniques of this disclosure.

As previously mentioned, recently, there has been a demand for 3D content for computer graphics, virtual reality, and communications, that has triggered a change in emphasis for the requirements. Many existing systems for constructing 3D models are built around specialized hardware that results in a high cost, which often cannot satisfy the requirements of these new applications. This need has stimulated the use of digital imaging facilities (e.g., cameras) for 3D reconstruction.

Currently, volume blocks (e.g., voxels) are often used to reconstruct a 3D scene from 2D images (e.g., stereo images obtained from a stereo camera). A voxel will be used herein as an example of blocks (e.g., 3D blocks). A voxel can represent a value on a regular grid in 3D space. As with pixels in a 2D bitmap, voxels themselves do not have their position (e.g., coordinates) explicitly encoded within their values. Instead, rendering systems infer the position of a voxel based upon its position relative to other voxels (e.g., its position in the data structure that makes up a single volumetric image).

3DR utilizes depth frames with an associated live camera pose estimate for scene reconstruction. In 3D surface reconstruction, the scene can be modeled as a 3D sparse volumetric representation (e.g., that can be referred to as a volume grid). The volume grid contains a set of voxel blocks that are indexed by their position in space with a sparse data representation (e.g., only storing blocks that surround an object and/or obstacle). For example, a room with a size of four meters (m) by four m by five m may be modeled with a volume grid having a total of 1.25 million (M) voxel blocks, where each voxel block has a four centimeter block dimension. In some examples, for this room, the occupied voxel blocks may only be about ten to fifteen percent.

FIG. 4 shows an example of a scene that has been modeled as a 3D sparse volumetric representation for 3DR. In particular, FIG. 4 is a diagram illustrating an example of a 3D surface reconstruction 400 of a scene modeled with an overlay of a volume grid containing voxel blocks. For 3DR, a camera (e.g., a stereo camera) may take photos of the scene from various different view points and angles. For example, a camera may take a photo of the scene when the camera is located at position P1. Once multiple photos have been taken of the scene, a 3D representation of the scene can be constructed by modeling the scene as a volume grid with 3D blocks (e.g., voxel blocks).

In one or more examples, an image (e.g., a photo) of a 3D block (e.g., voxel) located at point P2 within the scene may be taken by a camera (e.g., a stereo camera) located at point P1 with a certain camera pose (e.g., at a certain angle). The camera can capture both depth and color. From this image, it can be determined that there is an object located at point P2 with a certain depth and, as such, there is a surface. As such, it can be determined that there is an object that maps to this particular 3D block. An image of a 3D block located at point P3 within the scene may be taken by the same camera located at the point P1 with a different camera pose (e.g., with a different angle). From this image, it can be determined that there is an object located at point P3 with a certain depth and having a surface. As such, it can be determined that there is an object that maps to this particular 3D block (e.g., voxel). An integrate process can occur where all of the blocks within the scene are passed through an integrate function. The integrate function can determine depth information for each of the blocks from the depth frame and can update each block to indicate whether the block has a surface or not. The blocks that are determined to have a surface can then be updated with a color.

In one or more examples, the pose of the camera can indicate the location of the camera (e.g., which may be indicated by location coordinates X, Y) and the angle that the camera (e.g., which is the angle that the camera is positioned in for capturing the image). Each block (e.g., the block located at point P2) has a location (e.g., which may be indicated by location coordinates X, Y, Z). The pose of the camera and the location of each block can be used to map each block to world coordinates for the whole scene.

In one or more examples, to achieve fast multiple access to 3D blocks (e.g., voxels), instead of using a large memory lookup table, various different volume block representations may be used to index the blocks in the 3D scene to store data where the measurements are observed. Volume block representations that may be employed can include, but are not limited to, a hash map lookup, an octree, and a large blocks implementation.

FIG. 5 shows an example of a hash map lookup type of volume block representation. In particular, FIG. 5 is a diagram illustrating an example of a hash mapping function 500 for indexing voxel blocks 530 in a volume grid. In FIG. 5, a volume grid is shown with world coordinates 510. Also shown in FIG. 5 are a hash table 520 and voxel blocks 530. In one or more examples, a hash function can be used to map the integer world coordinates 510 into hash buckets 540 within the hash table 520. The hash buckets 540 can each store a small array of points to regular grid voxels blocks 530. Each voxel block 530 contains data that can be used for depth integration.

FIG. 6 is a diagram illustrating an example of a volume block (e.g., a voxel) 600. In FIG. 6, the voxel 600 is shown to have a block size of eight. For example, a 0.5 centimeter (cm) sample distance for an eight by eight by eight voxel can correspond to a four cm by four cm by four cm voxel.

In one or more examples, each voxel (e.g., voxel 600) can contain truncated signed distance function (TSDF) samples, a RGB, and a weight. TSDF is a function that measures the distance d of each pixel from the surface of an object to the camera. A voxel with a positive value for d can indicate that the voxel is located in front of a surface, a voxel with a negative value for d can indicate that the voxel is located inside (or behind) the surface, and a voxel with a zero value for d can indicate that the voxel is located on the surface. The distance d is truncated to [−1, 1], such that:

$\begin{array}{cc}\mathrm{tsdf}=\left\{\begin{array}{cc}-1,& \mathrm{if}d\le -\mathrm{ramp}\\ \frac{d}{\mathrm{ramp}},& \mathrm{if}-\mathrm{ramp}<d<\mathrm{ramp}\\ 1,& \mathrm{if}d\ge \mathrm{ramp}\end{array}\right\}& \left(\mathrm{equation}1\right)\end{array}$ $\begin{array}{cc}\mathrm{sample}.\mathrm{tsdf}=\left(\frac{\mathrm{sample}.\mathrm{weight}*\mathrm{sample}.\mathrm{tsdf}+\mathrm{tsdf}}{\mathrm{sample}.\mathrm{weight}+1}\right)& \left(\mathrm{equation}2\right)\end{array}$

A TSDF integration or fusion process can be employed that updates the TSDF values and weights with each new observation from the sensor (e.g., camera).

FIG. 7 is a diagram illustrating an example of a TSDF volume reconstruction 700. In FIG. 7, a voxel grid including a plurality of voxels is shown. A camera is shown to be obtaining images of a scene (e.g., person's face) from two different camera positions (e.g., camera position 1 710 and camera position 2 720). During operation for TSDF, for each new observation (e.g., image) from the camera (e.g., for each image taken by the camera at a different camera position), the distance (d) of a corresponding pixel of each voxel within the voxel grid can be obtained. The distance (d) can be subtracted from the distance of the voxel itself, and that result can be divided by a truncation threshold value threshold. The TSDF values (and color values) can be updated in the global memory. In FIG. 7, the voxels with positive values are shown to be located in front of the person's face, the voxels with negative values are shown to be located inside of the person's face, and the voxels with zero values are shown to be located on the surface of the person's face.

As previously mentioned, currently, end-to-end multi-view 3D reconstruction using deep learning has gained significant attention due to promising results shown by recent works. The main motivation of these works is to compute 2D visual features that are then lifted into a 3D volume of voxels. As there are multiple views of a scene, these works typically average the features to obtain a single embedding.

FIG. 8 is an example 3DR pipeline that averages the 2D visual features for 3DR. In particular, FIG. 8 is a diagram illustrating an example of a 3DR pipeline 800. In FIG. 8, a plurality of 2D image frames 820 are shown. In one or more examples, the image frames 820 may be obtained by a device (e.g., a camera, such as the image capture and processing system 100 of FIG. 1) that is moving throughout a scene of an environment and, as such, the device may capture continuous image frames of the scene and each image frame can include a different view of the scene.

During operation of the 3DR pipeline 800 of FIG. 8, the plurality of 2D image frames 820 of a scene of an environment can be passed through an image encoder 810 (e.g., a backbone feature extractor). The image encoder 810 can process the image frames 820 to output a set of deep learning features.

Depending upon the pose of the device (e.g., camera), each pixel in the image frames 820 can be back projected (or unprojected) into a 3D space, resulting in voxel features. Since there are multiple image frames 820 of the scene, there are multiple features (for each voxel) that are back projected into the 3D space. The multiple features for each voxel will need to be merged together into a single voxel feature for each voxel. One way to merge those features is to simply average out all of the features.

In one or more examples, the voxel features computed from the image frames 820 of the multiple views of the scene for a given voxel can be simply averaged. For example, assuming that there are N number of views of a scene (e.g., N number of image frames 820), for a given voxel v, a lifted feature may be obtained via a simple averaging (e.g., performed by one or more processors of the device) across views, such that a voxel feature is equal to:

$\begin{array}{cc}{f}_v=\frac{{\sum }_{i=1}^N{m}_i.f_i}{{\sum }_{i=1}^N{m}_i}& \left(\mathrm{equation}3\right)\end{array}$

where m_iϵ{0,1} is a visibility value (e.g., a mask value) showing whether the voxel is projected on the camera plane of view i, and f_i are the extracted 2D features of the 2D image frames 820.

A convolutional network 850 (e.g., including a stack of 3D convolutional layers) can then process the 3D features (e.g., voxel features) of the 3D space 840 to extract surfaces 860 of the scene from the visual features.

However, as previously mentioned, simply averaging the extracted features to obtain a single embedding can lead to over-smoothed surfaces in a 3DR. Various techniques can be performed to address such an issue, such as by utilizing expensive transformer layers to compute cross-attention amongst view features. Other techniques incur additional costs to first estimate the depth, which is then used to determine how the features should be aggregated. Therefore, systems and techniques for improved volumetric feature fusion for end-to-end 3DR can be useful.

In one or more aspects, the systems and techniques provide enhanced volumetric feature fusion based on geometric and similarity cues for end-to-end 3DR. In one or more examples, the systems and techniques use (e.g., instead of using a simple averaging of the extracted features) available geometric cues (e.g., the distance of voxels from camera centers, and angles between rays going through consecutive camera centers from a voxel) as well as a similarity of view features to learn adaptive scalars used as weights (e.g., normalized weights) for corresponding view features. The systems and techniques can ingest these cues and output appropriate weights (also referred to as normalized weights) that can better aggregate the multiple features unprojected to a voxel into a single voxel feature for the voxel, which can result in a higher 3D reconstruction quality, without introducing additional components with requiring significant computation. For instance, the systems and techniques can use a machine learning layer or component (e.g., multi-layer perceptron (MLP) network, a fully connected layer, and/or other architecture) to ingest the cues and generate the weights.

In one or more examples, the machine learning layer or component (e.g., the MLP, fully connected layer, or other component or layer) can be trained to generate weights (e.g., normalized weights, which may be scalars) for the respective unprojected features as output. The machine learning layer or component can be trained using information, such as 2D feature similarities across views and geometric cues (e.g., the distance of voxels from camera centers, and angles between rays going through consecutive camera centers from a voxel), as training data input to the machine learning component or layer. The final 3D reconstruction is based on the weighted average of the voxel features that are unprojected from the camera images.

In one or more aspects, the systems and techniques incorporate inherent geometric information that is available in multi-view 3D scene reconstruction without the need of any additional dedicated network and complex computing modules. As mentioned, in the existing mainstream solutions, voxel features computed from multiple views of a scene for a given voxel are simply averaged, resulting in an over-smoothed reconstruction. For instance, assuming there are multiple views of a scene, for a given voxel, a lifted feature (e.g., in existing mainstream solutions) is obtained via a simple averaging across views. The systems and techniques replace this simple averaging scheme by utilizing an adaptive weighted average mechanism that can determine how much each voxel feature should contribute, depending upon geometric cues (r, z_i, m_i) and visual similarity of computed features, where:

$\begin{array}{cc}{f}_v={\sum }_i{\alpha }_i\left({\stackrel{\_}{s}}_i,{\stackrel{\_}{r}}_i,z_i,m_i\right).f_i& \left(\mathrm{equation}4\right)\end{array}$ $\begin{array}{cc}0\le {\alpha }_i\left({\stackrel{\_}{s}}_i,{\stackrel{\_}{r}}_i,z_i,m_i\right)\le 1& \left(\mathrm{equation}5\right)\end{array}$ $\begin{array}{cc}{\sum }_i{\alpha }_i\left({\stackrel{\_}{s}}_i,{\stackrel{\_}{r}}_i,z_i,m_i\right)=1& \left(\mathrm{equation}6\right)\end{array}$ $\begin{array}{cc}\stackrel{\_}{s_i}={\left\{s_{i,j}\right\}}_{j=1,j\ne i}^N,s_{i,j}=f_i.f_j& \left(\mathrm{equation}7\right)\end{array}$

where z_i is the distance of the voxel to the view i's camera center, at are weights (e.g., normalized weights) that are positive values (e.g., scalar values) and sum together to equal one, m_iϵ{0,1} is a visibility value (e.g., a mask value) showing whether the voxel is projected on the camera plane of view i, f_i are the extracted 2D features, and s_i,j are similarity values between 2D features.

$\begin{array}{cc}\stackrel{\_}{r_i}={\left\{r_{i,j}\right\}}_{j=1,j\ne i}^N,r_{i,j}=\overrightarrow{r_i},\overrightarrow{r_j},\overrightarrow{r_i}& \left(\mathrm{equation}8\right)\end{array}$

where r_i is a ray going through the device (e.g., camera) center of view i from the voxel.

In one or more examples, the systems and techniques adaptively derive the weights of voxel features by incorporating geometric and similarity information from multiple views by using a verification network (e.g., implemented as a simple MLP) to ingest 2D feature similarities across views and geometric cues, and output the weights (α_i) for the respective features.

FIG. 9 shows an example of a verification network (e.g., an MLP) that may be employed by the systems and techniques. In particular, FIG. 9 is a diagram illustrating an example of a verification network 870 (e.g., an MLP) that may be employed for 3DR. As illustrated in FIG. 9, geometric information (e.g., distance 940 and consecutive view angle 950), similarity information (e.g., feature similarity 930), and mask information (e.g., validity mask 960) are input into the verification network 870.

The distance 940 is the distance from a voxel to a camera center (e.g., location of the device, such as an image sensor) for all views. For example, as shown in FIG. 9, the distance z_i (as an example of the distance 940) is the distance from a voxel 910 to the view i's 920a camera center (e.g., location of the device, such as an image sensor). Similarly, the distance (z_j) (as another example of the distance 940) is shown to be the distance from the voxel 910 to the view j's 920b camera center (e.g., location of the device or sensor). The feature similarity 930 (s_ij) is an indication of the visual similarity (or visual dissimilarity) of two (or more) features from different views for all of the different combinations of views. The consecutive view angle 950

$\left(\frac{\stackrel{\rightharpoonup }{r_i}.\stackrel{\rightharpoonup }{r_j}}{❘\stackrel{\rightharpoonup }{r_i}❘❘\stackrel{\rightharpoonup }{r_j}❘}\right)$

is the view angle between two rays (, ) for all of the different combinations of views.

The validity mask (m_i) is an indication of whether a voxel within the 3D scene can be observed by the different views. In one or more examples, the validity mask can be equal to a one (e.g., indicating that the voxel can be observed in the 3D scene for that particular view) or a zero (e.g., indicating that the voxel cannot be observed in the 3D scene for that particular view). In one or more examples, during inference of the verification network 870, based on camera pose information and voxel coordinates, views for each voxel projected (m_i) inside the camera plane are determined. If a voxel is not projected on any of camera plane views, the corresponding voxel's feature is set to zero in the validity mask, and is not processed by the verification network 870 (e.g., an MLP). Afterwards, the verification network 870 (e.g., an MLP) can process voxels with at least one visible view at a time. In some cases, batch processing may be performed, in which case all voxels can be processed at a same time.

The verification network 870 (e.g., an MLP) can process the various different inputted geometric information (e.g., distance 940 and consecutive view angle 950), the similarity information (e.g., feature similarity 930), and the mask information (e.g., validity mask 960), and can generate unnormalized weights γ, normalized weights (e.g., scalars) α_i, and voxel feature f_v (with multiple voxel features making up a feature volume). For example, the unnormalized weights γ can be represented as:

$\begin{array}{cc}\stackrel{\_}{\gamma }=\left[{\gamma }_1,\dots ,{\gamma }_9\right]& \left(\mathrm{equation}9\right)\end{array}$

Equation 9 is an example using nine frames (i=9) for illustrative purposes. Other numbers of frames can be used in other cases (e.g., depending on accuracy requirements, computation and/or memory constraints, etc.). An unnormalized weight γ_i for a frame i is generated by the verification network 870 corresponding to frame i. As described herein, there may be multiple frames taken from multiple views and, for each frame, one or more 2D features can be calculated. The 2D feature(s) can then be unprojected into one or more 3D locations (e.g., voxels) in a 3D space to generate voxel features. Due to multiple frames being captured (from multiple views), there may be multiple voxel features for each 3D location (e.g., for each voxel). The verification network 870 can compute the unnormalized weights γ_i, which are used to determine the normalized weights α_i (or adaptive weights) for use in averaging the voxel features. The verification network 870 can compute the unnormalized weights γ_i for each of the features. The unnormalized weights γ_i are not guaranteed to be positive and bounded between 0 and 1. The verification network 870 (or a separate component) can perform normalization (e.g., using equation 10 below) on the unnormalized weights γ_i to normalize the unnormalized weights γ_i and generate or obtain the normalized weights α_i.

The normalized weights α_i can be all positive values and sum together to equal one. The verification network 870 (e.g., an MLP) can derive the normalized weights by passing the unnormalized weights γ_i through a softmax layer (implementing a softmax algorithm) as follows:

$\begin{array}{cc}{\alpha }_i=\mathrm{softmax}\left(m_i{\gamma }_i\right)& \left(\mathrm{equation}10\right)\end{array}$

The verification network 870 (e.g., an MLP) can compute the voxel feature f_v, which is weighted based on the normalized weights α_i, using the following formula:

$\begin{array}{cc}{f}_v={\sum }_i{\alpha }_i{f}_i& \left(\mathrm{equation}11\right)\end{array}$

The features f_i are unprojected features. For example, as described above, 2D features can be extracted from image frames of a scene of an environment, with each image frame including a respective view of the scene. The 2D features are unprojected to a three-dimensional 3D space by assigning each 2D feature to a voxel (representing a 3D location in the scene of the environment), resulting in multiple features for each voxel. The weights α_i are adaptively calculated and are used to perform weighted averaging of the multiple features of a voxel to obtain a single voxel feature for the voxel. The resulting voxel features for the voxels of the 3D space are referred to as a feature volume. The verification network 870 (e.g., an MLP) can then output the voxel feature f_v computed by using equation 11. For example, the feature f_v and other voxel features for various voxels are used to perform 3D reconstruction for the feature volume.

In one or more examples, the systems and techniques may employ the pipeline 800 of FIG. 8, but can implement the verification network 870 (e.g., an MLP) within the pipeline 800 within a device (e.g., a camera, such as the image capture and processing system 100 of FIG. 1), and can substitute the calculation of voxel feature f_v as shown in equation 3 for the calculation of voxel features f_v as shown in equation 11. As such, referring back to FIG. 8, during operation of the 3DR pipeline 800 of FIG. 8, the plurality of 2D image frames 820 of a scene of an environment may be passed through the image encoder 810 (e.g., a backbone feature extractor). The image encoder 810 can process the image frames 820 to output a set of deep learning features. The verification network 870 (e.g., an MLP), implemented within the pipeline 800, can compute voxel features f_v, which are weighted using the normalized weights α_i according to equation 11.

After the verification network 870 (e.g., an MLP) of the device has determined the voxel features (e.g., using equation 11), one or more processors of the device can unproject the 2D features from a 2D space to a 3D space 840 based on the voxel features to generate or form the 3DR of the scene. A convolutional network 850 (e.g., including a stack of 3D convolutional layers) may then process the 3D features (e.g., voxel features) of the 3D space 840 to extract surfaces 860 of the scene from the visual features.

In one or more examples, an example of pseudo code for the processing of the verification network 870 (e.g., an MLP) to determine (e.g., compute) the voxel features f_v (e.g., by using equation 11) is as follows:

#Shape of mask3d: (num_views x D X H X W)
% the depth (D), height (H), and width (W) of the 3D space
mask = mask3d.view(num_views, −1).permute(1,0)
% the mask indicates whether a given voxel in a 3D scene can be observed by the
different views
valid_indices = ~(mask.sum(dim=1) == 0)
meta_data = get_similarity_and_geometric_cues(volume3d, valid_indices)
% feature similarity and geometric cues are in the form of metadata
view_weights = MLP(meta_data)
view_weights = view_weights * mask[valid_indices]
% weights (γi) are being multiplied by the masks
view_weights = nn.functional.softmax(view_weights, dim=1).unsqueeze(−1)
% passing through a softmax algorithm to produce weight scalars (αi)
feature3d = feature3d.view(num_views, feat_dim, −1).permute(2,0,1)
feature3d[valid_indices,:,:] *= view_weights
% determining the voxel features (fv) using the normalized weights (αi)

FIG. 10 shows a summary of performance results for 3DR comparing 3DR using a simple averaging of the 2D features to obtain the voxel features (e.g., using equation 3) with 3DR using a weighted average of the 2D features to obtain the voxel features (e.g., using equation 11). In particular, FIG. 10 is a table 1000 illustrating examples of 3DR performance results using a simple averaging of features (e.g., an existing setup) versus using a weighted average of features for the 3DR (e.g., geometric approach). The table 1000 shows 3DR performances for precision, recall, and F-score metrics. In FIG. 10, the table 1000 is shown to include a setting 1010 column, a precision 1020 column, a recall 1030 column, and a F-score 1040 column. As shown in the table 1000, using a weighted average of features for the 3DR (e.g., geometric approach) provides better 3DR performance than using a simple averaging of features for the 3DR (e.g., existing approach) across all of the performance metrics.

Not only does 3DR using a weighted average of the 2D features to obtain the voxel features (e.g., using equation 11) provide a quantitative improvement in performance over 3DR using a simple averaging of the 2D features to obtain the voxel features (e.g., using equation 3), but 3DR using a weighted average of the 2D features to obtain the voxel features (e.g., using equation 11) also provides a qualitative improvement in performance over 3DR using a simple averaging of the 2D features to obtain the voxel features (e.g., using equation 3). FIG. 11 is a diagram illustrating examples 1100 of visual 3DR results using a simple averaging of features versus using a weighted average of features for the 3DR. In FIG. 11, the third row (e.g., bottom row) shows ground truth meshes 1130 of 3D scenes. The first row (e.g., top row) shows 3DR of the ground truth meshes 1130 by using a baseline network 1110 (e.g., using a simple averaging of features). The second row (e.g., middle row) shows 3DR of the ground truth meshes 1130 by using a weighting network 1120 (e.g., using a weighted average of features). In FIG. 11, the 3DR using the baseline network 1110 is shown to have more smoothed-out features than the 3DR using the weighting network 1120.

In one or more examples, the systems and techniques may also utilize additional information, such as semantic labels, depth prediction, depth uncertainties, and 3D scans to enhance the multi-view feature aggregation. In some examples, the weights (e.g., normalized weights at) for the features may be further based on semantic labels, depth prediction, depth uncertainties, and/or 3D scans.

In some examples, the systems and techniques can also use neighboring voxel's feature similarity across views to improve the robustness of the weighing mechanism. In some examples, one or more voxel features can be grouped together to generate or form a group of voxel features (e.g., a patch of voxel features), based on the one or more voxel features having a visual similarity across different views of the scene. The unprojection of the plurality of image frames from the 2D space onto the 3D space can then be based on the group of voxel features (e.g., the patch of voxel features). This grouping of voxel features can create a consensus and can make the adaptive weighing mechanism more robust. This neighboring process can be more efficient for processing than an individual processing.

In one or more examples, the systems and techniques can employ a lookup table (LUT) to map the geometric cues to weights (e.g., the normalized weights α_i) after training the MLP to reduce the computation complexity. In one or more examples, determining the weights (e.g., the normalized weights α_i) for the plurality of features can be further based on a LUT mapping the geometric information to weights of the plurality of weights.

FIG. 12 is a flow chart illustrating an example of a process 1200 for enhanced volumetric feature fusion based on geometric and similarity cues for end-to-end 3DR. The process 1200 can be performed by a computing device or apparatus (e.g., image capture and processing system 100 of FIG. 1) or by a component or system (e.g., one or more chipsets, one or more processors such as one or more CPUs, DSPs, NPUs, NSPs, microcontrollers, ASICs, FPGAs, programmable logic devices, discrete gates or transistor logic components, discrete hardware components, etc., an ML system such as a neural network model, any combination thereof, and/or other component or system) of the computing device or apparatus. The computing device or apparatus may be a vehicle or component or system of a vehicle, a mobile device (e.g., a mobile phone), a network-connected wearable such as a watch, an extended reality (XR) device (e.g., a virtual reality (VR) device, augmented reality (AR) device, and/or mixed reality (MR) device), or other type of computing device. The operations of the process 1200 may be implemented as software components that are executed and run on one or more processors (e.g., processor 1310 of FIG. 13 or other processor(s)). Further, the transmission and/or reception of signals by the device in the process 1200 may be enabled, for example, by one or more antennas and/or one or more transceivers (e.g., wireless transceiver(s)).

At block 1210, the computing device (or component thereof) can extract a plurality of two-dimensional (2D) features from a plurality of image frames of a scene. Each image frame of the plurality of image frames comprises a respective view of the scene.

At block 1220, the computing device (or component thereof) can unproject the plurality of 2D features from a 2D space onto a three-dimensional (3D) space to obtain a plurality of 3D features.

At block 1230, the computing device (or component thereof) can determine a plurality of weights (e.g., normalized weights α_i) for the plurality of 3D features based on geometric information and visual similarity information for the plurality of image frames. Each weight of the plurality of weights is associated with a respective feature 3D of the plurality of 3D features. In some cases, the computing device (or component thereof) can determine unnormalized weights (e.g., the unnormalized weights γ, such as that shown in equation 9). The unnormalized weights can be used to generate the plurality of weights (e.g., the normalized weights α_i according to equation 10 using a softmax function or other normalization function). Once normalized, each weight of the plurality of weights is a positive value. In some cases, once normalized, a sum of all weights of the plurality of weights is equal to one.

In some aspects, the visual similarity information is based on a visual similarity (e.g., feature similarity 930 of FIG. 9) of two or more 2D features of the plurality of 2D features of the plurality of 2D features from different images frames of the plurality of image frames. It is noted that unprojection of the 2D features into the 3D space may not change the content of the features (e.g., the unprojection assigns the 2D features to a 3D grid made up of voxels). Thus, the visual similarity computation can be the same either in 2D or 3D, as the content is the same and only the ordering differs.

In some cases, the geometric information is based on a plurality of distances. For instance, each distance of the plurality of distances can be between a respective voxel associated with a respective voxel feature of the plurality of voxel features to a location of a respective image sensor (e.g., a camera). In one illustrative example, the the distance z_i shown in FIG. 9 (as an example of the distance 940) is the distance from a voxel 910 to the view i's 920a camera center (e.g., location of the device, such as an image sensor), and the distance (zj) (as another example of the distance 940) is the distance from the voxel 910 to the view j's 920b camera center (e.g., location of the device or sensor). In some aspects, the geometric information is based on a plurality of view angles (e.g., the view angle 950 of FIG. 9). For instance, each view angle of the plurality of view angles can be between two respective rays each associated with a respective image frame of the plurality of image frames. For example, as described with respect to FIG. 9, the consecutive view angle 950

$\left(\frac{\stackrel{\rightharpoonup }{r_i}.\stackrel{\rightharpoonup }{r_j}}{❘\stackrel{\rightharpoonup }{r_i}❘❘\stackrel{\rightharpoonup }{r_j}❘}\right)$

is the view angle between two rays (, ) for all of the different combinations of views.

In some aspects, the computing device (or component thereof) can determine the plurality of weights for the plurality of 3D features further based on at least one of semantic labels, depth predictions, depth uncertainties, or 3D scans associated with the plurality of 2D features. In some cases, the computing device (or component thereof) can determine the plurality of weights for the plurality of 3D features further based on a lookup table. For instance, the lookup table can map the geometric information to weights of the plurality of weights.

At block 1240, the computing device (or component thereof) can determine a plurality of voxel features (e.g., the voxel features f_v, such as that determined using equation 11) for the plurality of image frames based on the plurality of weights for the plurality of 3D features. In some aspects, the computing device (or component thereof) can group one or more voxel features of the plurality of voxel features together to generate a group of voxel features based on the one or more voxel features having a visual similarity across different views of the scene. In such aspects, unprojecting the plurality of 2D features from the 2D space onto the 3D space can be based on the group of voxel features. The grouping of the voxel features can create a consensus and can make the adaptive weighing mechanism more robust.

At block 1250, the computing device (or component thereof) can generate the 3DR of the scene based on the plurality of voxel features.

In some examples, the process 1200 may be performed by one or more computing devices or apparatuses. In some illustrative examples, the process 1200 can be performed by image capture and processing system 100 of FIG. 1 and/or one or more computing devices or systems (e.g., the computing system 1300 of FIG. 13). In some cases, such a computing device or apparatus may include a processor, microprocessor, microcomputer, or other component of a device that is configured to carry out the steps of the process 1200. Such computing device may further include a network interface configured to communicate data.

The components of the computing device can be implemented in circuitry. For example, the components can include and/or can be implemented using electronic circuits or other electronic hardware, which can include one or more programmable electronic circuits (e.g., microprocessors, graphics processing units (GPUs), digital signal processors (DSPs), central processing units (CPUs), and/or other suitable electronic circuits), and/or can include and/or be implemented using computer software, firmware, or any combination thereof, to perform the various operations described herein. The computing device may further include a display (as an example of the output device or in addition to the output device), a network interface configured to communicate and/or receive the data, any combination thereof, and/or other component(s). The network interface may be configured to communicate and/or receive Internet Protocol (IP) based data or other type of data.

The process 1200 is illustrated as a logical flow diagram, the operations of which represent a sequence of operations that can be implemented in hardware, computer instructions, or a combination thereof. In the context of computer instructions, the operations represent computer-executable instructions stored on one or more computer-readable storage media that, when executed by one or more processors, perform the recited operations. Generally, computer-executable instructions include routines, programs, objects, components, data structures, and the like that perform particular functions or implement particular data types. The order in which the operations are described is not intended to be construed as a limitation, and any number of the described operations can be combined in any order and/or in parallel to implement the processes.

Additionally, the process 1200 may be performed under the control of one or more computer systems configured with executable instructions and may be implemented as code (e.g., executable instructions, one or more computer programs, or one or more applications) executing collectively on one or more processors, by hardware, or combinations thereof. As noted above, the code may be stored on a computer-readable or machine-readable storage medium, for example, in the form of a computer program comprising a plurality of instructions executable by one or more processors. The computer-readable or machine-readable storage medium may be non-transitory.

FIG. 13 is a block diagram illustrating an example of a computing system 1300, which may be employed by the disclosed system for enhanced volumetric feature fusion based on geometric and similarity cues for end-to-end 3DR. In particular, FIG. 13 illustrates an example of computing system 1300, which can be for example any computing device making up internal computing system, a remote computing system, a camera, or any component thereof in which the components of the system are in communication with each other using connection 1305. Connection 1305 can be a physical connection using a bus, or a direct connection into processor 1310, such as in a chipset architecture. Connection 1305 can also be a virtual connection, networked connection, or logical connection.

In some aspects, computing system 1300 is a distributed system in which the functions described in this disclosure can be distributed within a datacenter, multiple data centers, a peer network, etc. In some aspects, one or more of the described system components represents many such components each performing some or all of the function for which the component is described. In some aspects, the components can be physical or virtual devices.

Example system 1300 includes at least one processing unit (CPU or processor) 1310 and connection 1305 that communicatively couples various system components including system memory 1315, such as read-only memory (ROM) 1320 and random access memory (RAM) 1325 to processor 1310. Computing system 1300 can include a cache 1312 of high-speed memory connected directly with, in close proximity to, or integrated as part of processor 1310.

Processor 1310 can include any general purpose processor and a hardware service or software service, such as services 1332, 1334, and 1336 stored in storage device 1330, configured to control processor 1310 as well as a special-purpose processor where software instructions are incorporated into the actual processor design. Processor 1310 may essentially be a completely self-contained computing system, containing multiple cores or processors, a bus, memory controller, cache, etc. A multi-core processor may be symmetric or asymmetric.

To enable user interaction, computing system 1300 includes an input device 1345, which can represent any number of input mechanisms, such as a microphone for speech, a touch-sensitive screen for gesture or graphical input, keyboard, mouse, motion input, speech, etc. Computing system 1300 can also include output device 1335, which can be one or more of a number of output mechanisms. In some instances, multimodal systems can enable a user to provide multiple types of input/output to communicate with computing system 1300.

Computing system 1300 can include communications interface 1340, which can generally govern and manage the user input and system output. The communication interface may perform or facilitate receipt and/or transmission wired or wireless communications using wired and/or wireless transceivers, including those making use of an audio jack/plug, a microphone jack/plug, a universal serial bus (USB) port/plug, an Apple™ Lightning™ port/plug, an Ethernet port/plug, a fiber optic port/plug, a proprietary wired port/plug, 3G, 4G, 5G and/or other cellular data network wireless signal transfer, a Bluetooth™ wireless signal transfer, a Bluetooth™ low energy (BLE) wireless signal transfer, an IBEACON™ wireless signal transfer, a radio-frequency identification (RFID) wireless signal transfer, near-field communications (NFC) wireless signal transfer, dedicated short range communication (DSRC) wireless signal transfer, 802.11 Wi-Fi wireless signal transfer, wireless local area network (WLAN) signal transfer, Visible Light Communication (VLC), Worldwide Interoperability for Microwave Access (WiMAX), Infrared (IR) communication wireless signal transfer, Public Switched Telephone Network (PSTN) signal transfer, Integrated Services Digital Network (ISDN) signal transfer, ad-hoc network signal transfer, radio wave signal transfer, microwave signal transfer, infrared signal transfer, visible light signal transfer, ultraviolet light signal transfer, wireless signal transfer along the electromagnetic spectrum, or some combination thereof.

The communications interface 1340 may also include one or more range sensors (e.g., LIDAR sensors, laser range finders, RF radars, ultrasonic sensors, and infrared (IR) sensors) configured to collect data and provide measurements to processor 1310, whereby processor 1310 can be configured to perform determinations and calculations needed to obtain various measurements for the one or more range sensors. In some examples, the measurements can include time of flight, wavelengths, azimuth angle, elevation angle, range, linear velocity and/or angular velocity, or any combination thereof. The communications interface 1340 may also include one or more Global Navigation Satellite System (GNSS) receivers or transceivers that are used to determine a location of the computing system 1300 based on receipt of one or more signals from one or more satellites associated with one or more GNSS systems. GNSS systems include, but are not limited to, the US-based GPS, the Russia-based Global Navigation Satellite System (GLONASS), the China-based BeiDou Navigation Satellite System (BDS), and the Europe-based Galileo GNSS. There is no restriction on operating on any particular hardware arrangement, and therefore the basic features here may easily be substituted for improved hardware or firmware arrangements as they are developed.

Storage device 1330 can be a non-volatile and/or non-transitory and/or computer-readable memory device and can be a hard disk or other types of computer readable media which can store data that are accessible by a computer, such as magnetic cassettes, flash memory cards, solid state memory devices, digital versatile disks, cartridges, a floppy disk, a flexible disk, a hard disk, magnetic tape, a magnetic strip/stripe, any other magnetic storage medium, flash memory, memristor memory, any other solid-state memory, a compact disc read only memory (CD-ROM) optical disc, a rewritable compact disc (CD) optical disc, digital video disk (DVD) optical disc, a blu-ray disc (BDD) optical disc, a holographic optical disk, another optical medium, a secure digital (SD) card, a micro secure digital (microSD) card, a Memory Stick® card, a smartcard chip, a EMV chip, a subscriber identity module (SIM) card, a mini/micro/nano/pico SIM card, another integrated circuit (IC) chip/card, random access memory (RAM), static RAM (SRAM), dynamic RAM (DRAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash EPROM (FLASHEPROM), cache memory (e.g., Level 1 (L1) cache, Level 2 (L2) cache, Level 3 (L3) cache, Level 4 (L4) cache, Level 5 (L5) cache, or other (L #) cache), resistive random-access memory (RRAM/ReRAM), phase change memory (PCM), spin transfer torque RAM (STT-RAM), another memory chip or cartridge, and/or a combination thereof.

The storage device 1330 can include software services, servers, services, etc., that when the code that defines such software is executed by the processor 1310, it causes the system to perform a function. In some aspects, a hardware service that performs a particular function can include the software component stored in a computer-readable medium in connection with the necessary hardware components, such as processor 1310, connection 1305, output device 1335, etc., to carry out the function. The term “computer-readable medium” includes, but is not limited to, portable or non-portable storage devices, optical storage devices, and various other mediums capable of storing, containing, or carrying instruction(s) and/or data. A computer-readable medium may include a non-transitory medium in which data can be stored and that does not include carrier waves and/or transitory electronic signals propagating wirelessly or over wired connections. Examples of a non-transitory medium may include, but are not limited to, a magnetic disk or tape, optical storage media such as compact disk (CD) or digital versatile disk (DVD), flash memory, memory or memory devices. A computer-readable medium may have stored thereon code and/or machine-executable instructions that may represent a procedure, a function, a subprogram, a program, a routine, a subroutine, a module, a software package, a class, or any combination of instructions, data structures, or program statements. A code segment may be coupled to another code segment or a hardware circuit by passing and/or receiving information, data, arguments, parameters, or memory contents. Information, arguments, parameters, data, etc. may be passed, forwarded, or transmitted via any suitable means including memory sharing, message passing, token passing, network transmission, or the like.

Specific details are provided in the description above to provide a thorough understanding of the aspects and examples provided herein, but those skilled in the art will recognize that the application is not limited thereto. Thus, while illustrative aspects of the application have been described in detail herein, it is to be understood that the inventive concepts may be otherwise variously embodied and employed, and that the appended claims are intended to be construed to include such variations, except as limited by the prior art. Various features and aspects of the above-described application may be used individually or jointly. Further, aspects can be utilized in any number of environments and applications beyond those described herein without departing from the broader scope of the specification. The specification and drawings are, accordingly, to be regarded as illustrative rather than restrictive. For the purposes of illustration, methods were described in a particular order. It should be appreciated that in alternate aspects, the methods may be performed in a different order than that described.

For clarity of explanation, in some instances the present technology may be presented as including individual functional blocks comprising devices, device components, steps or routines in a method embodied in software, or combinations of hardware and software. Additional components may be used other than those shown in the figures and/or described herein. For example, circuits, systems, networks, processes, and other components may be shown as components in block diagram form in order not to obscure the aspects in unnecessary detail. In other instances, well-known circuits, processes, algorithms, structures, and techniques may be shown without unnecessary detail in order to avoid obscuring the aspects.

Further, those of skill in the art will appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the aspects disclosed herein may be implemented as electronic hardware, computer software, or combinations of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.

Individual aspects may be described above as a process or method which is depicted as a flowchart, a flow diagram, a data flow diagram, a structure diagram, or a block diagram. Although a flowchart may describe the operations as a sequential process, many of the operations can be performed in parallel or concurrently. In addition, the order of the operations may be re-arranged. A process is terminated when its operations are completed, but could have additional steps not included in a figure. A process may correspond to a method, a function, a procedure, a subroutine, a subprogram, etc. When a process corresponds to a function, its termination can correspond to a return of the function to the calling function or the main function.

Processes and methods according to the above-described examples can be implemented using computer-executable instructions that are stored or otherwise available from computer-readable media. Such instructions can include, for example, instructions and data which cause or otherwise configure a general purpose computer, special purpose computer, or a processing device to perform a certain function or group of functions. Portions of computer resources used can be accessible over a network. The computer executable instructions may be, for example, binaries, intermediate format instructions such as assembly language, firmware, source code. Examples of computer-readable media that may be used to store instructions, information used, and/or information created during methods according to described examples include magnetic or optical disks, flash memory, USB devices provided with non-volatile memory, networked storage devices, and so on.

In some aspects the computer-readable storage devices, mediums, and memories can include a cable or wireless signal containing a bitstream and the like. However, when mentioned, non-transitory computer-readable storage media expressly exclude media such as energy, carrier signals, electromagnetic waves, and signals per se.

Those of skill in the art will appreciate that information and signals may be represented using any of a variety of different technologies and techniques. For example, data, instructions, commands, information, signals, bits, symbols, and chips that may be referenced throughout the above description may be represented by voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields or particles, or any combination thereof, in some cases depending in part on the particular application, in part on the desired design, in part on the corresponding technology, etc.

The various illustrative logical blocks, modules, and circuits described in connection with the aspects disclosed herein may be implemented or performed using hardware, software, firmware, middleware, microcode, hardware description languages, or any combination thereof, and can take any of a variety of form factors. When implemented in software, firmware, middleware, or microcode, the program code or code segments to perform the necessary tasks (e.g., a computer-program product) may be stored in a computer-readable or machine-readable medium. A processor(s) may perform the necessary tasks. Examples of form factors include laptops, smart phones, mobile phones, tablet devices or other small form factor personal computers, personal digital assistants, rackmount devices, standalone devices, and so on. Functionality described herein also can be embodied in peripherals or add-in cards. Such functionality can also be implemented on a circuit board among different chips or different processes executing in a single device, by way of further example.

The instructions, media for conveying such instructions, computing resources for executing them, and other structures for supporting such computing resources are example means for providing the functions described in the disclosure.

The techniques described herein may also be implemented in electronic hardware, computer software, firmware, or any combination thereof. Such techniques may be implemented in any of a variety of devices such as general purposes computers, wireless communication device handsets, or integrated circuit devices having multiple uses including application in wireless communication device handsets and other devices. Any features described as modules or components may be implemented together in an integrated logic device or separately as discrete but interoperable logic devices. If implemented in software, the techniques may be realized at least in part by a computer-readable data storage medium comprising program code including instructions that, when executed, performs one or more of the methods, algorithms, and/or operations described above. The computer-readable data storage medium may form part of a computer program product, which may include packaging materials. The computer-readable medium may comprise memory or data storage media, such as random access memory (RAM) such as synchronous dynamic random access memory (SDRAM), read-only memory (ROM), non-volatile random access memory (NVRAM), electrically erasable programmable read-only memory (EEPROM), FLASH memory, magnetic or optical data storage media, and the like. The techniques additionally, or alternatively, may be realized at least in part by a computer-readable communication medium that carries or communicates program code in the form of instructions or data structures and that can be accessed, read, and/or executed by a computer, such as propagated signals or waves.

The program code may be executed by a processor, which may include one or more processors, such as one or more digital signal processors (DSPs), general purpose microprocessors, an application specific integrated circuits (ASICs), field programmable logic arrays (FPGAs), or other equivalent integrated or discrete logic circuitry. Such a processor may be configured to perform any of the techniques described in this disclosure. A general-purpose processor may be a microprocessor; but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing devices, e.g., a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration. Accordingly, the term “processor,” as used herein may refer to any of the foregoing structure, any combination of the foregoing structure, or any other structure or apparatus suitable for implementation of the techniques described herein.

One of ordinary skill will appreciate that the less than (“<”) and greater than (“>”) symbols or terminology used herein can be replaced with less than or equal to (“≤”) and greater than or equal to (“≥”) symbols, respectively, without departing from the scope of this description.

Where components are described as being “configured to” perform certain operations, such configuration can be accomplished, for example, by designing electronic circuits or other hardware to perform the operation, by programming programmable electronic circuits (e.g., microprocessors, or other suitable electronic circuits) to perform the operation, or any combination thereof.

The phrase “coupled to” or “communicatively coupled to” refers to any component that is physically connected to another component either directly or indirectly, and/or any component that is in communication with another component (e.g., connected to the other component over a wired or wireless connection, and/or other suitable communication interface) either directly or indirectly.

Claim language or other language reciting “at least one of” a set and/or “one or more” of a set indicates that one member of the set or multiple members of the set (in any combination) satisfy the claim. For example, claim language reciting “at least one of A and B” or “at least one of A or B” means A, B, or A and B. In another example, claim language reciting “at least one of A, B, and C” or “at least one of A, B, or C” means A, B, C, or A and B, or A and C, or B and C, A and B and C, or any duplicate information or data (e.g., A and A, B and B, C and C, A and A and B, and so on), or any other ordering, duplication, or combination of A, B, and C. The language “at least one of” a set and/or “one or more” of a set does not limit the set to the items listed in the set. For example, claim language reciting “at least one of A and B” or “at least one of A or B” may mean A, B, or A and B, and may additionally include items not listed in the set of A and B. The phrases “at least one” and “one or more” are used interchangeably herein.

Claim language or other language reciting “at least one processor configured to,” “at least one processor being configured to,” “one or more processors configured to,” “one or more processors being configured to,” or the like indicates that one processor or multiple processors (in any combination) can perform the associated operation(s). For example, claim language reciting “at least one processor configured to: X, Y, and Z” means a single processor can be used to perform operations X, Y, and Z; or that multiple processors are each tasked with a certain subset of operations X, Y, and Z such that together the multiple processors perform X, Y, and Z; or that a group of multiple processors work together to perform operations X, Y, and Z. In another example, claim language reciting “at least one processor configured to: X, Y, and Z” can mean that any single processor may only perform at least a subset of operations X, Y, and Z.

Where reference is made to one or more elements performing functions (e.g., steps of a method), one element may perform all functions, or more than one element may collectively perform the functions. When more than one element collectively performs the functions, each function need not be performed by each of those elements (e.g., different functions may be performed by different elements) and/or each function need not be performed in whole by only one element (e.g., different elements may perform different sub-functions of a function). Similarly, where reference is made to one or more elements configured to cause another element (e.g., an apparatus) to perform functions, one element may be configured to cause the other element to perform all functions, or more than one element may collectively be configured to cause the other element to perform the functions.

Where reference is made to an entity (e.g., any entity or device described herein) performing functions or being configured to perform functions (e.g., steps of a method), the entity may be configured to cause one or more elements (individually or collectively) to perform the functions. The one or more components of the entity may include at least one memory, at least one processor, at least one communication interface, another component configured to perform one or more (or all) of the functions, and/or any combination thereof. Where reference to the entity performing functions, the entity may be configured to cause one component to perform all functions, or to cause more than one component to collectively perform the functions. When the entity is configured to cause more than one component to collectively perform the functions, each function need not be performed by each of those components (e.g., different functions may be performed by different components) and/or each function need not be performed in whole by only one component (e.g., different components may perform different sub-functions of a function).

The various illustrative logical blocks, modules, engines, circuits, and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, firmware, or combinations thereof. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, engines, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.

The techniques described herein may also be implemented in electronic hardware, computer software, firmware, or any combination thereof. Such techniques may be implemented in any of a variety of devices such as general purposes computers, wireless communication device handsets, or integrated circuit devices having multiple uses including application in wireless communication device handsets and other devices. Any features described as engines, modules, or components may be implemented together in an integrated logic device or separately as discrete but interoperable logic devices. If implemented in software, the techniques may be realized at least in part by a computer-readable data storage medium comprising program code including instructions that, when executed, performs one or more of the methods described above. The computer-readable data storage medium may form part of a computer program product, which may include packaging materials. The computer-readable medium may comprise memory or data storage media, such as random access memory (RAM) such as synchronous dynamic random access memory (SDRAM), read-only memory (ROM), non-volatile random access memory (NVRAM), electrically erasable programmable read-only memory (EEPROM), FLASH memory, magnetic or optical data storage media, and the like. The techniques additionally, or alternatively, may be realized at least in part by a computer-readable communication medium that carries or communicates program code in the form of instructions or data structures and that can be accessed, read, and/or executed by a computer, such as propagated signals or waves.

The program code may be executed by a processor, which may include one or more processors, such as one or more digital signal processors (DSPs), general purpose microprocessors, an application specific integrated circuits (ASICs), field programmable logic arrays (FPGAs), or other equivalent integrated or discrete logic circuitry. Such a processor may be configured to perform any of the techniques described in this disclosure. A general purpose processor may be a microprocessor; but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing devices, e.g., a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration. Accordingly, the term “processor,” as used herein may refer to any of the foregoing structure, any combination of the foregoing structure, or any other structure or apparatus suitable for implementation of the techniques described herein. In addition, in some aspects, the functionality described herein may be provided within dedicated software modules or hardware modules configured for encoding and decoding, or incorporated in a combined video encoder-decoder (CODEC).

Illustrative aspects of the disclosure include:

Aspect 1. An apparatus for three-dimensional reconstruction (3DR) of a scene, the apparatus comprising: at least one memory; and at least one processor coupled to the at least one memory and configured to: extract a plurality of two-dimensional (2D) features from a plurality of image frames of a scene, wherein each image frame of the plurality of image frames comprises a respective view of the scene; unproject the plurality of 2D features from a 2D space onto a three-dimensional (3D) space to obtain a plurality of 3D features; determine a plurality of weights for the plurality of 3D features based on geometric information and visual similarity information for the plurality of image frames, wherein each weight of the plurality of weights is associated with a respective feature 3D of the plurality of 3D features; determine a plurality of voxel features for the plurality of image frames based on the plurality of weights for the plurality of 3D features; and generate the 3DR of the scene based on the plurality of voxel features.

Aspect 2. The apparatus of Aspect 1, wherein the visual similarity information is based on a visual similarity of two or more 2D features of the plurality of 2D features of the plurality of 2D features from different images frames of the plurality of image frames.

Aspect 3. The apparatus of any of Aspects 1 or 2, wherein the geometric information is based on a plurality of distances, wherein each distance of the plurality of distances is between a respective voxel associated with a respective voxel feature of the plurality of voxel features to a location of a respective image sensor.

Aspect 4. The apparatus of Aspect 3, wherein the respective image sensor is a camera.

Aspect 5. The apparatus of any of Aspects 1 to 4, wherein the geometric information is based on a plurality of view angles, wherein each view angle of the plurality of view angles is between two respective rays each associated with a respective image frame of the plurality of image frames.

Aspect 6. The apparatus of any of Aspects 1 to 5, wherein the at least one processor is configured to determine the plurality of weights for the plurality of 3D features further based on at least one of semantic labels, depth predictions, depth uncertainties, or 3D scans associated with the plurality of 2D features.

Aspect 7. The apparatus of any of Aspects 1 to 6, wherein the at least one processor is configured to group one or more voxel features of the plurality of voxel features together to generate a group of voxel features based on the one or more voxel features having a visual similarity across different views of the scene, wherein unprojecting the plurality of 2D features from the 2D space onto the 3D space is based on the group of voxel features.

Aspect 8. The apparatus of any of Aspects 1 to 7, wherein the at least one processor is configured to determine the plurality of weights for the plurality of 3D features further based on a lookup table, the lookup table mapping the geometric information to weights of the plurality of weights.

Aspect 9. The apparatus of any of Aspects 1 to 8, wherein each weight of the plurality of weights is a positive value.

Aspect 10. The apparatus of any of Aspects 1 to 9, wherein a sum of all weights of the plurality of weights is equal to one.

Aspect 11. A method for three-dimensional reconstruction (3DR) of a scene, the method comprising: extracting a plurality of two-dimensional (2D) features from a plurality of image frames of a scene, wherein each image frame of the plurality of image frames comprises a respective view of the scene; unprojecting the plurality of 2D features from a 2D space onto a three-dimensional (3D) space to obtain a plurality of 3D features; determining a plurality of weights for the plurality of 3D features based on geometric information and visual similarity information for the plurality of image frames, wherein each weight of the plurality of weights is associated with a respective feature 3D of the plurality of 3D features; determining a plurality of voxel features for the plurality of image frames based on the plurality of weights for the plurality of 3D features; and generating the 3DR of the scene based on the plurality of voxel features.

Aspect 12. The method of Aspect 11, wherein the visual similarity information is based on a visual similarity of two or more 2D features of the plurality of 2D features from different images frames of the plurality of image frames.

Aspect 13. The method of any of Aspects 11 or 12, wherein the geometric information is based on a plurality of distances, wherein each distance of the plurality of distances is between a respective voxel associated with a respective voxel feature of the plurality of voxel features to a location of a respective image sensor.

Aspect 14. The method of Aspect 13, wherein the respective image sensor is a camera.

Aspect 15. The method of any of Aspects 11 to 14, wherein the geometric information is based on a plurality of view angles, wherein each view angle of the plurality of view angles is between two respective rays each associated with a respective image frame of the plurality of image frames.

Aspect 16. The method of Aspects 11 to 15, wherein determining the plurality of weights for the plurality of 3D features is further based on at least one of semantic labels, depth predictions, depth uncertainties, or 3D scans associated with the plurality of 2D features.

Aspect 17. The method of Aspects 11 to 16, further comprising grouping one or more voxel features of the plurality of voxel features together to generate a group of voxel features based on the one or more voxel features having a visual similarity across different views of the scene, wherein unprojecting the plurality of 2D features from the 2D space onto the 3D space is based on the group of voxel features.

Aspect 18. The method of Aspects 11 to 17, wherein determining the plurality of weights for the plurality of 3D features is further based on a lookup table, the lookup table mapping the geometric information to weights of the plurality of weights.

Aspect 19. The method of Aspects 11 to 18, wherein each weight of the plurality of weights is a positive value.

Aspect 20. The method of Aspects 11 to 19, wherein a sum of all weights of the plurality of weights is equal to one.

Aspect 21. A non-transitory computer-readable medium having stored thereon instructions that, when executed by one or more processors, cause the one or more processors to perform operations according to any of Aspects 11 to 20.

Aspect 22. An apparatus for three-dimensional reconstruction (3DR) of a scene, the apparatus including one or more means for performing operations according to any of Aspects 11 to 20.

The previous description is provided to enable any person skilled in the art to practice the various aspects described herein. Various modifications to these aspects will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other aspects. Thus, the claims are not intended to be limited to the aspects shown herein, but is to be accorded the full scope consistent with the language claims, wherein reference to an element in the singular is not intended to mean “one and only one” unless specifically so stated, but rather “one or more.”

Claims

1. An apparatus for three-dimensional reconstruction (3DR) of a scene, the apparatus comprising: at least one memory; andat least one processor coupled to the at least one memory and configured to: extract a plurality of two-dimensional (2D) features from a plurality of image frames of a scene, wherein each image frame of the plurality of image frames comprises a respective view of the scene;unproject the plurality of 2D features from a 2D space onto a three-dimensional (3D) space to obtain a plurality of 3D features;determine a plurality of weights for the plurality of 3D features based on geometric information and visual similarity information for the plurality of image frames, wherein each weight of the plurality of weights is associated with a respective feature 3D of the plurality of 3D features;determine a plurality of voxel features for the plurality of image frames based on the plurality of weights for the plurality of 3D features; andgenerate the 3DR of the scene based on the plurality of voxel features.

2. The apparatus of claim 1, wherein the visual similarity information is based on a visual similarity of two or more 2D features of the plurality of 2D features of the plurality of 2D features from different images frames of the plurality of image frames.

3. The apparatus of claim 1, wherein the geometric information is based on a plurality of distances, wherein each distance of the plurality of distances is between a respective voxel associated with a respective voxel feature of the plurality of voxel features to a location of a respective image sensor.

4. The apparatus of claim 3, wherein the respective image sensor is a camera.

5. The apparatus of claim 1, wherein the geometric information is based on a plurality of view angles, wherein each view angle of the plurality of view angles is between two respective rays each associated with a respective image frame of the plurality of image frames.

6. The apparatus of claim 1, wherein the at least one processor is configured to determine the plurality of weights for the plurality of 3D features further based on at least one of semantic labels, depth predictions, depth uncertainties, or 3D scans associated with the plurality of 2D features.

7. The apparatus of claim 1, wherein the at least one processor is configured to group one or more voxel features of the plurality of voxel features together to generate a group of voxel features based on the one or more voxel features having a visual similarity across different views of the scene, wherein unprojecting the plurality of 2D features from the 2D space onto the 3D space is based on the group of voxel features.

8. The apparatus of claim 1, wherein the at least one processor is configured to determine the plurality of weights for the plurality of 3D features further based on a lookup table, the lookup table mapping the geometric information to weights of the plurality of weights.

9. The apparatus of claim 1, wherein each weight of the plurality of weights is a positive value.

10. The apparatus of claim 1, wherein a sum of all weights of the plurality of weights is equal to one.

11. A method for three-dimensional reconstruction (3DR) of a scene, the method comprising: extracting a plurality of two-dimensional (2D) features from a plurality of image frames of a scene, wherein each image frame of the plurality of image frames comprises a respective view of the scene;unprojecting the plurality of 2D features from a 2D space onto a three-dimensional (3D) space to obtain a plurality of 3D features;determining a plurality of weights for the plurality of 3D features based on geometric information and visual similarity information for the plurality of image frames, wherein each weight of the plurality of weights is associated with a respective feature 3D of the plurality of 3D features;determining a plurality of voxel features for the plurality of image frames based on the plurality of weights for the plurality of 3D features; andgenerating the 3DR of the scene based on the plurality of voxel features.

12. The method of claim 11, wherein the visual similarity information is based on a visual similarity of two or more 2D features of the plurality of 2D features from different images frames of the plurality of image frames.

13. The method of claim 11, wherein the geometric information is based on a plurality of distances, wherein each distance of the plurality of distances is between a respective voxel associated with a respective voxel feature of the plurality of voxel features to a location of a respective image sensor.

14. The method of claim 13, wherein the respective image sensor is a camera.

15. The method of claim 11, wherein the geometric information is based on a plurality of view angles, wherein each view angle of the plurality of view angles is between two respective rays each associated with a respective image frame of the plurality of image frames.

16. The method of claim 11, wherein determining the plurality of weights for the plurality of 3D features is further based on at least one of semantic labels, depth predictions, depth uncertainties, or 3D scans associated with the plurality of 2D features.

17. The method of claim 11, further comprising grouping one or more voxel features of the plurality of voxel features together to generate a group of voxel features based on the one or more voxel features having a visual similarity across different views of the scene, wherein unprojecting the plurality of 2D features from the 2D space onto the 3D space is based on the group of voxel features.

18. The method of claim 11, wherein determining the plurality of weights for the plurality of 3D features is further based on a lookup table, the lookup table mapping the geometric information to weights of the plurality of weights.

19. The method of claim 11, wherein each weight of the plurality of weights is a positive value.

20. The method of claim 11, wherein a sum of all weights of the plurality of weights is equal to one.