Patent Application
20240282013
The present embodiments generally relate to a method and an apparatus for point cloud compression and processing.
The Point Cloud (PC) data format is a universal data format across several business domains, e.g., from autonomous driving, robotics, augmented reality/virtual reality (AR/VR), civil engineering, computer graphics, to the animation/movie industry. 3D LiDAR (Light Detection and Ranging) sensors have been deployed in self-driving cars, and affordable LiDAR sensors are released from Velodyne Velabit, Apple iPad Pro 2020 and Intel RealSense LiDAR camera L515. With advances in sensing technologies, 3D point cloud data becomes more practical than ever and is expected to be an ultimate enabler in the applications discussed herein.
According to an embodiment, a method for decoding point cloud data is provided, comprising: accessing a data array with samples on a regular grid, wherein each sample in said data array indicates a position of a point in a point cloud; and reconstructing said point cloud responsive to said data array. The data array may be decoded by a decoder associated with a neural network-based autoencoder, or an image or video decoder. In addition, a codeword that provides an initial representation of said point cloud may be decoded, wherein said point cloud is reconstructed further responsive to said codeword.
According to another embodiment, a method for encoding point cloud data is provided, comprising: generating a codeword, by a first neural network-based module, which provides a representation of an input point cloud associated with said point cloud data; reconstructing a first point cloud, by a second neural network-based module, based on said codeword and a grid; and generating a data array with samples on said grid, wherein each sample in said data array indicates a position of a point in said input point cloud, based on said reconstructed first point cloud, said grid, and said input point cloud. The data array or codeword may be compressed. The data array can be encoded by an encoder associated with a neural network-based autoencoder, or an image or video encoder.
According to another embodiment, an apparatus for decoding point cloud data is presented, comprising one or more processors, wherein said one or more processors are configured to access a data array with samples on a regular grid, wherein each sample in said data array indicates a position of a point in a point cloud; and reconstruct said point cloud responsive to said data array. The data array may be decoded by a decoder associated with a neural network-based autoencoder, or an image or video decoder. In addition, a codeword that provides an initial representation of said point cloud may be decoded, wherein said point cloud is reconstructed further responsive to said codeword. The apparatus may further include at least one memory coupled to said said more or more processors.
According to another embodiment, an apparatus for encoding point cloud data is presented, comprising one or more processors, wherein said one or more processors are configured to generate a codeword, by a first neural network-based module, which provides a representation of an input point cloud associated with said point cloud data; reconstruct a first point cloud, by a second neural network-based module, based on said codeword and a grid; and generate a data array with samples on said grid, wherein each sample in said data array indicates a position of a point in said input point cloud, based on said reconstructed first point cloud, said grid, and said input point cloud. The data array or codeword may be compressed. The data array can be encoded by an encoder associated with a neural network-based autoencoder, or an image or video encoder.
One or more embodiments also provide a computer program comprising instructions which when executed by one or more processors cause the one or more processors to perform the encoding method or decoding method according to any of the embodiments described above. One or more of the present embodiments also provide a computer readable storage medium having stored thereon instructions for encoding or decoding point cloud data according to the methods described above.
One or more embodiments also provide a computer readable storage medium having stored thereon a bitstream generated according to the methods described above. One or more embodiments also provide a method and apparatus for transmitting or receiving the bitstream generated according to the methods described above.
The system 100 includes at least one processor 110 configured to execute instructions loaded therein for implementing, for example, the various aspects described in this application. Processor 110 may include embedded memory, input output interface, and various other circuitries as known in the art. The system 100 includes at least one memory 120 (e.g., a volatile memory device, and/or a non-volatile memory device). System 100 includes a storage device 140, which may include non-volatile memory and/or volatile memory, including, but not limited to, EEPROM, ROM, PROM, RAM, DRAM, SRAM, flash, magnetic disk drive, and/or optical disk drive. The storage device 140 may include an internal storage device, an attached storage device, and/or a network accessible storage device, as non-limiting examples.
System 100 includes an encoder/decoder module 130 configured, for example, to process data to provide an encoded video or decoded video, and the encoder/decoder module 130 may include its own processor and memory. The encoder/decoder module 130 represents module(s) that may be included in a device to perform the encoding and/or decoding functions. As is known, a device may include one or both of the encoding and decoding modules. Additionally, encoder/decoder module 130 may be implemented as a separate element of system 100 or may be incorporated within processor 110 as a combination of hardware and software as known to those skilled in the art.
Program code to be loaded onto processor 110 or encoder/decoder 130 to perform the various aspects described in this application may be stored in storage device 140 and subsequently loaded onto memory 120 for execution by processor 110. In accordance with various embodiments, one or more of processor 110, memory 120, storage device 140, and encoder/decoder module 130 may store one or more of various items during the performance of the processes described in this application. Such stored items may include, but are not limited to, the input video, the decoded video or portions of the decoded video, the bitstream, matrices, variables, and intermediate or final results from the processing of equations, formulas, operations, and operational logic.
In several embodiments, memory inside of the processor 110 and/or the encoder/decoder module 130 is used to store instructions and to provide working memory for processing that is needed during encoding or decoding. In other embodiments, however, a memory external to the processing device (for example, the processing device may be either the processor 110 or the encoder/decoder module 130) is used for one or more of these functions. The external memory may be the memory 120 and/or the storage device 140, for example, a dynamic volatile memory and/or a non-volatile flash memory. In several embodiments, an external non-volatile flash memory is used to store the operating system of a television. In at least one embodiment, a fast external dynamic volatile memory such as a RAM is used as working memory for video coding and decoding operations, such as for MPEG-2, HEVC, or VVC.
The input to the elements of system 100 may be provided through various input devices as indicated in block 105. Such input devices include, but are not limited to, (i) an RF portion that receives an RF signal transmitted, for example, over the air by a broadcaster, (ii) a Composite input terminal, (iii) a USB input terminal, and/or (iv) an HDMI input terminal.
In various embodiments, the input devices of block 105 have associated respective input processing elements as known in the art. For example, the RF portion may be associated with elements suitable for (i) selecting a desired frequency (also referred to as selecting a signal, or band-limiting a signal to a band of frequencies), (ii) down converting the selected signal, (iii) band-limiting again to a narrower band of frequencies to select (for example) a signal frequency band which may be referred to as a channel in certain embodiments, (iv) demodulating the down converted and band-limited signal, (v) performing error correction, and (vi) demultiplexing to select the desired stream of data packets. The RF portion of various embodiments includes one or more elements to perform these functions, for example, frequency selectors, signal selectors, band-limiters, channel selectors, filters, downconverters, demodulators, error correctors, and demultiplexers. The RF portion may include a tuner that performs various of these functions, including, for example, down converting the received signal to a lower frequency (for example, an intermediate frequency or a near-baseband frequency) or to baseband. In one set-top box embodiment, the RF portion and its associated input processing element receives an RF signal transmitted over a wired (for example, cable) medium, and performs frequency selection by filtering, down converting, and filtering again to a desired frequency band. Various embodiments rearrange the order of the above-described (and other) elements, remove some of these elements, and/or add other elements performing similar or different functions. Adding elements may include inserting elements in between existing elements, for example, inserting amplifiers and an analog-to-digital converter. In various embodiments, the RF portion includes an antenna.
Additionally, the USB and/or HDMI terminals may include respective interface processors for connecting system 100 to other electronic devices across USB and/or HDMI connections. It is to be understood that various aspects of input processing, for example, Reed-Solomon error correction, may be implemented, for example, within a separate input processing IC or within processor 110 as necessary. Similarly, aspects of USB or HDMI interface processing may be implemented within separate interface ICs or within processor 110 as necessary. The demodulated, error corrected, and demultiplexed stream is provided to various processing elements, including, for example, processor 110, and encoder/decoder 130 operating in combination with the memory and storage elements to process the datastream as necessary for presentation on an output device.
Various elements of system 100 may be provided within an integrated housing, Within the integrated housing, the various elements may be interconnected and transmit data therebetween using suitable connection arrangement 115, for example, an internal bus as known in the art, including the I2C bus, wiring, and printed circuit boards.
The system 100 includes communication interface 150 that enables communication with other devices via communication channel 190. The communication interface 150 may include, but is not limited to, a transceiver configured to transmit and to receive data over communication channel 190. The communication interface 150 may include, but is not limited to, a modem or network card and the communication channel 190 may be implemented, for example, within a wired and/or a wireless medium.
Data is streamed to the system 100, in various embodiments, using a Wi-Fi network such as IEEE 802.11. The Wi-Fi signal of these embodiments is received over the communications channel 190 and the communications interface 150 which are adapted for Wi-Fi communications. The communications channel 190 of these embodiments is typically connected to an access point or router that provides access to outside networks including the Internet for allowing streaming applications and other over-the-top communications. Other embodiments provide streamed data to the system 100 using a set-top box that delivers the data over the HDMI connection of the input block 105. Still other embodiments provide streamed data to the system 100 using the RF connection of the input block 105.
The system 100 may provide an output signal to various output devices, including a display 165, speakers 175, and other peripheral devices 185. The other peripheral devices 185 include, in various examples of embodiments, one or more of a stand-alone DVR, a disk player, a stereo system, a lighting system, and other devices that provide a function based on the output of the system 100. In various embodiments, control signals are communicated between the system 100 and the display 165, speakers 175, or other peripheral devices 185 using signaling such as AV. Link, CEC, or other communications protocols that enable device-to-device control with or without user intervention. The output devices may be communicatively coupled to system 100 via dedicated connections through respective interfaces 160, 170, and 180. Alternatively, the output devices may be connected to system 100 using the communications channel 190 via the communications interface 150. The display 165 and speakers 175 may be integrated in a single unit with the other components of system 100 in an electronic device, for example, a television. In various embodiments, the display interface 160 includes a display driver, for example, a timing controller (T Con) chip.
The display 165 and speaker 175 may alternatively be separate from one or more of the other components, for example, if the RF portion of input 105 is part of a separate set-top box. In various embodiments in which the display 165 and speakers 175 are external components, the output signal may be provided via dedicated output connections, including, for example, HDMI ports, USB ports, or COMP outputs.
It is contemplated that point cloud data may consume a large portion of network traffic, e.g., among connected cars over 5G network, and immersive communications (VR/AR). Efficient representation formats are necessary for point cloud understanding and communication. In particular, raw point cloud data need to be properly organized and processed for the purposes of world modeling and sensing. Compression on raw point clouds is essential when storage and transmission of the data are required in the related scenarios.
Furthermore, point clouds may represent a sequential scan of the same scene, which contains multiple moving objects. They are called dynamic point clouds as compared to static point clouds captured from a static scene or static objects. Dynamic point clouds are typically organized into frames, with different frames being captured at different times. Dynamic point clouds may require the processing and compression to be in real-time or with low delay.
The automotive industry and autonomous car are domains in which point clouds may be used. Autonomous cars should be able to “probe” their environment to make good driving decisions based on the reality of their immediate surroundings. Typical sensors like LiDARs produce (dynamic) point clouds that are used by the perception engine. These point clouds are not intended to be viewed by human eyes and they are typically sparse, not necessarily colored, and dynamic with a high frequency of capture. They may have other attributes like the reflectance ratio provided by the LiDAR as this attribute is indicative of the material of the sensed object and may help in making a decision.
Virtual Reality (VR) and immersive worlds are foreseen by many as the future of 2D flat video. For VR and immersive worlds, a viewer is immersed in an environment all around the viewer, as opposed to standard TV where the viewer can only look at the virtual world in front of the viewer. There are several gradations in the immersivity depending on the freedom of the viewer in the environment. Point cloud is a good format candidate to distribute VR worlds. The point cloud for use in VR may be static or dynamic and are typically of average size, for example, no more than millions of points at a time.
Point clouds may also be used for various purposes such as culture heritage/buildings in which objects like statues or buildings are scanned in 3D in order to share the spatial configuration of the object without sending or visiting the object. Also, point clouds may also be used to ensure preservation of the knowledge of the object in case the object may be destroyed, for instance, a temple by an earthquake. Such point clouds are typically static, colored, and huge.
Another use case is in topography and cartography in which using 3D representations, maps are not limited to the plane and may include the relief Google Maps is a good example of 3D maps but uses meshes instead of point clouds. Nevertheless, point clouds may be a suitable data format for 3D maps and such point clouds are typically static, colored, and huge.
World modeling and sensing via point clouds could be a useful technology to allow machines to gain knowledge about the 3D world around them for the applications discussed herein.
3D point cloud data are essentially discrete samples on the surfaces of objects or scenes. To fully represent the real world with point samples, in practice it requires a huge number of points. For instance, a typical VR immersive scene contains millions of points, while point clouds typically contain hundreds of millions of points. Therefore, the processing of such large-scale point clouds is computationally expensive, especially for consumer devices, e.g., smartphone, tablet, and automotive navigation system, that have limited computational power.
In order to perform processing or inference on a point cloud, efficient storage methodologies are needed. To store and process an input point cloud with affordable computational cost, one solution is to down-sample the point cloud first, where the down-sampled point cloud summarizes the geometry of the input point cloud while having much fewer points. The down-sampled point cloud is then fed to the subsequent machine task for further consumption. However, further reduction in storage space can be achieved by converting the raw point cloud data (original or down-sampled) into a bitstream through entropy coding techniques for lossless compression. Better entropy models result in a smaller bitstream and hence more efficient compression. Additionally, the entropy models can also be paired with downstream tasks which allow the entropy encoder to maintain the task-specific information while compressing.
In addition to lossless coding, many scenarios seek lossy coding for significantly improved compression ratio while maintaining the induced distortion under certain quality levels.
Point cloud compression (PCC) refers to the problem of succinctly representing the surface manifold of the object(s) contained within a point cloud. Several fronts in regards to point cloud compression have been explored and can broadly be categorized into the following categories: PCC in the input domain, PCC in the primitive domain, PCC in the transform domain, and finally PCC via entropy coding. PCC in the input domain refers to down-sampling the raw point cloud by choosing or generating novel key points that are representative of the underlying surface manifold. Although several learned (deep learning-based) and classical machine learning techniques exist in this area, many PCCs in the input domain are only suitable for dense point clouds as the network is restricted to do regular convolution. For PCC in the primitive domain, key points primitives (regular geometric 2D/3D shapes) are generated that aim to closely follow the underlying object manifold. PCC in the transform domain refers to the case when the raw point cloud data is first transformed into another domain via classical methods and then the transformed representation in the new domain is compressed to obtain more efficient compression. Though some work can be interpreted as transformation, it is non-trivial to have them been applied for a compression system. Finally, in the case of PCC via entropy coding, either the raw point cloud data or another (trivially obtained) representation of the point cloud is entropy coded via either adaptive learning-based or classical methods.
Generally, the raw point cloud data obtained from sensing modalities contains huge number of unorganized points that need to be stored efficiently. However, the irregularities and sparsity of point cloud data make it difficult for compression.
The present application relates to transform-based PCC, since we propose to organize an irregular point cloud as a 2D image which can then be compressed with popular transform-based approaches such as JPEG, MPEG AVC/HEVC/VVC. In one embodiment, based on a neural network, we unfold an input point cloud onto a regular, organized grid structure to achieve efficient coding of the point cloud data.
FoldingNet is an autoencoder developed in the context of high-level computer vision problems, e.g., classification/segmentation, via unsupervised learning.
A 2D grid structure, which is an image based on a pre-defined sampling pattern on a 2D surface, is introduced into the inverse transform (decoder) of FoldingNet. In one embodiment, it is a two-channel image representing 2D coordinates regularly sampled in a square region. In another embodiment, it consists of coordinates regularly sampled on a 2D sphere.
Given an original point cloud PCo (M points), the transforming module PN (210) generates a codeword CW in a latent space. Here, a latent space refers to an abstract multi-dimensional feature space, and the codeword CW is a feature vector representing the point cloud. Typically, the codeword can provide a high-level description or a rough representation of the point cloud.
Then the module FN (220) performs an inverse transforming process. It takes the 2D grid as an input in addition to the codeword CW, and endeavors to reconstruct another point cloud PC1 which is close to the input PC0. The 2D grid contains N=W×H grid points, where W and H are the width and height of the grid image. In
The design of FoldingNet promotes the high-level representability of codeword CW. With CW alone, it is intractable to reconstruct fine details in a point cloud, i.e., hard to reconstruct each individual point accurately. Hence, compressing the codeword CW alone from FoldingNet will not solve the point cloud compression problem where pointwise distortion also matters in addition to high-level representability.
To address this challenge, the UnfoldingOperator is proposed. The UnfoldingOperator directly embeds (or unfolds) a raw input point cloud onto a regular image. Compared to the codeword CW, the image representation from the UnfoldingOperator contains detailed information of the point cloud. Moreover, since the image data format is dense and organized, it is more suitable for down-stream processing, such as point cloud compression.
In one embodiment, to unfold the input point cloud, it is proposed to utilize the FN module to facilitate a newly proposed UF module as shown in
Specifically, the UF module (320) takes PC0, the reconstruction PC′ and its corresponding 2D grid as inputs. It first matches each point in PC′ to a point in PC0. Suppose point P′ from PC′ is matched with point P0 from PC0, then the 3D coordinate of P0 is put onto the 2D-grid position associated with P′ (i.e., the 2D grid position that is mapped to point P′ in the reconstructed point cloud PC′). In this way, a 3-channel image on the 2D grid, XYZ, is constructed as the unfolding output. In a more general use case, not only the 3D coordinates but also other point attributes, such as color, normal, reflectance, etc., are put onto the 2D-grid. In that case, a K-channel image on the 2D grid (still denoted as XYZ without loss of generality) can be constructed as the unfolding output. Generally, the unfolding output can be seen as a data array with samples on the 2D grid, each sample includes K components indicating point attributes, for example, 3D position, color, normal, and/or reflectance.
Note that the FN module guides the mapping between PC0 and PC′ through its output PC′. It helps the XYZ image to preserve smoothness, i.e., two neighboring points in PC0 are likely to be neighbors in the XYZ image.
The unfolded XYZ image is not only useful for the compression tasks, but also makes it possible to bring neural network based approaches from image domain to point cloud domain, because a non-organized point cloud is able to be represented in a three-channel image format. In particular, many neural network based methods in image domain rely on the pixel arranging format, that cannot be directly applied for point cloud tasks. However, by virtue of the proposed UF module, the generated XYZ image can be directly fed into neural networks that process regular images, e.g., convolutional neural networks (CNNs).
The proposed UnfoldingOperator is related to the TearingTransform (and TearingCompression) as described in a commonly owned U.S. Provisional Application No. 63/181,270, entitled “Learning-Based Point Cloud Compression Via Tearing Transform” (Attorney Docket Number 2021PF00130). TearingTransform (and TearingCompression) also aims to reconstruct an input point cloud with fine details. A diagram of the TearingTransform is shown in
The UV image in TearingNet is introduced and updated within the inverse transformer (decoder). The UV image is defined based on a 2D sampling grid at certain resolution. In one embodiment, the two channels in the UV image represents coordinates in the 2D space. In another embodiment, they represent the coordinate offsets with respect to their default positions in the 2D space.
The codeword is replicated N times and the resulting Nx512 matrix (610) is concatenated with an Nx2 matrix (620) that contains the N grid points in the 2D grid. The result of the concatenation is a matrix of size Nx514 (630), which is fed to the first series of shared MLP layers (640) to output a matrix of size Nx3 (650). Then the replicated codewords are concatenated to the Nx3 output (650) to form a matrix of size Nx515 (660), which is fed into the second series of shared MLP layers (670). The final output PC′ (680) is the reconstructed point cloud represented by Nx3 matrix, where N is the number of points in the output point cloud PC′.
For each point P′ in PC′, a corresponding point P0 is first identified, for example, via nearest neighbor search, using the NN module (710). We note that point P′ corresponds to a 2D point (u′, v′) in the 2D grid as described before for the FN module. In other words, the 2D position (u′, v′) can be indexed/retrieved via P′. Then the UF module puts the coordinate of P0=(x0, y0, z0) onto the 2D-grid position (u′, v′) associated with P′, i.e., XYZ (u′, v′, 1)=x0, XYZ (u′, v′, 2)=y0, and XYZ (u′, v′, 3)=z0. Performing this operation for each point in PC′ leads to the UF output, XYZ, which is a 3-channel image.
In one embodiment, except for the 3D coordinates P0=(x0, y0, z0), other features associated with the 3D point P0, for example, but not limited to, color (RGB), normal vector, and reflectance, are also put onto the 2D-grid position (u′, v′) to form the XYZ image. For example, XYZ (u′, v′, 4)=R, XYZ (u′, v′, 5)=G, XYZ (u′, v′, 6)=B, XYZ (u′, v′, 7)=nx, XYZ (u′, v′, 8)=ny, XYZ (u′, v, 9)=nz, and XYZ (u′, v′, 10)=r, where (R, G, B) is the color at P0, (nx, ny, nz) is the normal vector at P0 and r is the reflectance at P0. In this embodiment, the XYZ image contains the 3D coordinates as well as other point features, namely, the XYZ image is a K-channel image where K>3.
In the above, the examples of the PN, FN, and UF modules are described. It should be noted that these modules can use different network structures or configurations. For example, the MLP dimensions may be adjusted according to the complexity of practical scenarios, or more sets of MLPs can be used. Generally, any network structure that meets the input/output requirements can be used.
In the previous embodiment, the XYZ image serves as the only representation of PC0. In this embodiment, an input point cloud is decomposed into codeword CW and another 3-channel image, denoted as ΔXYZ. The codeword CW is used to reconstruct a rough shape of the point cloud, while ΔXYZ is to represent fine details on top of the rough reconstruction.
In the differential UnfoldingOperator, the UF module computes the residual between PC0 and PC′ then put the residual on the image ΔXYZ. Therefore, CW and ΔXYZ together represent PC0, and ΔXYZ is used to compensate the error.
Similar to the XYZ image representation, the ΔXYZ can also include more information. Generally, the differential unfolding output can be seen as a data array with samples on the 2D grid, each sample includes K components where the first three components contain Δx, Δy and Δz (the residual of the 3D positions), and the rest components include other attributes, for example, the RGB color, the normal vector and the reflectance.
In this embodiment, it is proposed to apply the UnfoldingOperator in a learning-based point cloud compression system. The overall diagram of the proposed UnfoldingCompression is shown in
In one embodiment, the compression of the XYZ image can be based on state-of-the-art image/video compression methods, e.g., JPEG, MPEG AVC/HEVC/VVC. As described before, the information indicative of point positions ({X, Y, Z} or {ΔX, ΔY, ΔZ}) may be arranged into a 3-channel image with each position-indicative parameter carried on one of the three channels. Quantization (1110) is performed before compression, for example, in order to convert the floating numbers in the XYZ image to a data format used by the 2D video encoder. Also, without any adjustments, both the XYZ image and the ΔXYZ image may have negative values, we can normalize them and make their dynamic range fall into a pre-defined interval before sending them to the codec. In one embodiment, we first compute the minimum and the maximum values of each channel of the XYZ (or ΔXYZ) image, denoted by mink and maxk where k ranges from 1 to K. Then we normalize each channel of the XYZ image to the range, for example, [0, 255], before feeding it to the codec. In this case, the minimum and the maximum values mink and maxk of each channel need to be sent as metadata to facilitate the decoding. Note that the minimum and the maximum values may be floating point numbers and can take negative values.
In another embodiment, the compression can be neural network-based methods, such as a variational autoencoder based on a factorized prior model or a scale hyperprior model, which approximates the quantization operation by adding uniform noise. It generates a differentiable bitrate RXYZ for end-to-end training. Since the neighboring samples in the XYZ or ΔXYZ images usually represent neighboring points in the original point cloud, usually there is strong correlation between neighboring samples. Thus, we anticipate that the XYZ and the ΔXYZ images would be coded efficiently with (unmodified) standard image and video codecs.
On the decoder side, provided a bitstream of the XYZ image as input, the XYZ image is decoded (1120). The reconstruction PC1 is simply the 3D points on the XYZ image. In another embodiment where metadata is also received, the reconstruction also relies on the received metadata. For example, when the minimum and the maximum values mink and maxk of each channel is received on the decoder side, they are used to have each channel of the reconstruction PC1 scaled back to their original range of values.
In another embodiment, the differential UnfoldingOperator is applied for PCC, as shown in
In this embodiment, the compression (1230) of the ΔXYZ image can be similarly done as the XYZ image in UnfoldingCompression (
Provided bitstreams on the codeword CW′ and ΔXYZ image as inputs, the decoder first reconstructs (1240, 1250) CW′ and ΔXYZ from the coded symbols, then fed CW′ to the FN module (1260), which takes input of both codeword CW and 2D grid to reconstruct a preliminary point cloud PC′. Then the residual ΔXYZ is added (1270) back to PC′ to obtain the reconstruction PC1. Note that the decoder uses decompression methods that correspond to the compression methods used to generate the codeword CW and grid.
In the example
Working with TearingTransform or TearingCompression
In this embodiment, it is proposed to apply the UnfoldingOperator with the TearingTransform to form a point cloud reconstruction system. In general, the TearingTransform works well when the points of input point clouds exhibit a fairly regular distribution. However, if the input point clouds are too sparse for the TN module in the TearingTransform (
With the same rationale, it is proposed to apply the UnfoldingCompression with the TearingCompression to form a point cloud compression system. Specifically, when TearingCompression fails to produce high-quality decoded point cloud due to the failure of the TN module, the system switches to the proposed UF module to guarantee accurate reconstructions.
The UnfoldingOperator in
The training method of UnfoldingOperator could be extended as follows to train UnfoldingCompression in
In the first stage, a FoldingNet (PN and FN) is first trained in the same way described earlier. In the second stage, end-to-end training on UnfoldingCompression will be conducted. The PN and FN modules are initialized with the parameters learned from the first stage. Then a rate-distortion loss is used for training. Specifically, the reconstruction quality, measured by CD or EMD between PC0 and PC1, is regularized by the bitrate of the XYZ bitstream, as shown below:
where dPC
For differential UnfoldingCompression, an additional bitstream for the latent code CW is needed. Then the training loss becomes:
where RCW is the CW bitstream, and μ is an additional coefficient to balance its importance.
Often a compressed point cloud is not solely to be viewed by human eyes, but also for machine-oriented tasks, e.g., classification or segmentation. In such scenarios, our proposal is further extended for performing machine tasks.
The diagram of applying UnfoldingCompression for machine tasks is presented in
The proposed differential UnfoldingCompression can also be extended for machine tasks, as shown in
In the above embodiments, the input point cloud to be encoded is a full point cloud frame. In another embodiment, it is proposed to first partition full point cloud frames into smaller point cloud blocks. Then the point cloud blocks are fed to the proposed UnfoldingOperator or UnfoldingCompression as inputs to limit the complexity required to process the input point clouds. In an embodiment, to compress the XYZ (or ΔXYZ) images of the small point cloud blocks with state-of-the-art image/video compression methods, e.g., JPEG, MPEG AVC/HiEVC/VVC, the XYZ (or ΔXYZ) images are tiled into a large image. The tiling can be based on the Morton order of the associated 3D block, or another pre-defined order. The tiling can also be arranged to make the induced large image more friendly to the downstream image/video codec, e.g., by minimizing the differences across neighboring image blocks. If a dynamic point cloud is fed to the system, then we get a sequence of tiled XYZ (or ΔXYZ) images which can be coded with existing video codecs.
Various numeric values are used in the present application. The specific values are for example purposes and the aspects described are not limited to these specific values.
The implementations and aspects described herein may be implemented in, for example, a method or a process, an apparatus, a software program, a data stream, or a signal. Even if only discussed in the context of a single form of implementation (for example, discussed only as a method), the implementation of features discussed may also be implemented in other forms (for example, an apparatus or program). An apparatus may be implemented in, for example, appropriate hardware, software, and firmware. The methods may be implemented in, for example, an apparatus, for example, a processor, which refers to processing devices in general, including, for example, a computer, a microprocessor, an integrated circuit, or a programmable logic device. Processors also include communication devices, for example, computers, cell phones, portable/personal digital assistants (“PDAs”), and other devices that facilitate communication of information between end-users.
Reference to “one embodiment” or “an embodiment” or “one implementation” or “an implementation”, as well as other variations thereof, means that a particular feature, structure, characteristic, and so forth described in connection with the embodiment is included in at least one embodiment. Thus, the appearances of the phrase “in one embodiment” or “in an embodiment” or “in one implementation” or “in an implementation”, as well any other variations, appearing in various places throughout this application are not necessarily all referring to the same embodiment.
Additionally, this application may refer to “determining” various pieces of information. Determining the information may include one or more of, for example, estimating the information, calculating the information, predicting the information, or retrieving the information from memory.
Further, this application may refer to “accessing” various pieces of information. Accessing the information may include one or more of, for example, receiving the information, retrieving the information (for example, from memory), storing the information, moving the information, copying the information, calculating the information, determining the information, predicting the information, or estimating the information.
Additionally, this application may refer to “receiving” various pieces of information. Receiving is, as with “accessing”, intended to be a broad term. Receiving the information may include one or more of, for example, accessing the information, or retrieving the information (for example, from memory). Further, “receiving” is typically involved, in one way or another, during operations, for example, storing the information, processing the information, transmitting the information, moving the information, copying the information, erasing the information, calculating the information, determining the information, predicting the information, or estimating the information.
It is to be appreciated that the use of any of the following “/”, “and/or”, and “at least one of”, for example, in the cases of “A/B”, “A and/or B” and “at least one of A and B”, is intended to encompass the selection of the first listed option (A) only, or the selection of the second listed option (B) only, or the selection of both options (A and B). As a further example, in the cases of “A, B, and/or C” and “at least one of A, B, and C”, such phrasing is intended to encompass the selection of the first listed option (A) only, or the selection of the second listed option (B) only, or the selection of the third listed option (C) only, or the selection of the first and the second listed options (A and B) only, or the selection of the first and third listed options (A and C) only, or the selection of the second and third listed options (B and C) only, or the selection of all three options (A and B and C). This may be extended, as is clear to one of ordinary skill in this and related arts, for as many items as are listed.
As will be evident to one of ordinary skill in the art, implementations may produce a variety of signals formatted to carry information that may be, for example, stored or transmitted. The information may include, for example, instructions for performing a method, or data produced by one of the described implementations. For example, a signal may be formatted to carry the bitstream of a described embodiment. Such a signal may be formatted, for example, as an electromagnetic wave (for example, using a radio frequency portion of spectrum) or as a baseband signal. The formatting may include, for example, encoding a data stream and modulating a carrier with the encoded data stream. The information that the signal carries may be, for example, analog or digital information. The signal may be transmitted over a variety of different wired or wireless links, as is known. The signal may be stored on a processor-readable medium.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/US2022/034184 | 6/20/2022 | WO |
| Number | Date | Country | |
|---|---|---|---|
| 63213654 | Jun 2021 | US |
