Patent Application
20250240423
The disclosed embodiments relate generally to video coding, including but not limited to systems and methods of selecting transform kernels.
Digital video is supported by a variety of electronic devices, such as digital televisions, laptop or desktop computers, tablet computers, digital cameras, digital recording devices, digital media players, video gaming consoles, smart phones, video teleconferencing devices, video streaming devices, etc. The electronic devices transmit and receive or otherwise communicate digital video data across a communication network, and/or store the digital video data on a storage device. Due to a limited bandwidth capacity of the communication network and limited memory resources of the storage device, video coding may be used to compress the video data according to one or more video coding standards before it is communicated or stored. The video coding can be performed by hardware and/or software on an electronic/client device or a server providing a cloud service.
Video coding generally utilizes prediction methods (e.g., inter-prediction, intra-prediction, or the like) that take advantage of redundancy inherent in the video data. Video coding aims to compress video data into a form that uses a lower bit rate, while avoiding or minimizing degradations to video quality. Multiple video codec standards have been developed. For example, High-Efficiency Video Coding (HEVC/H.265) is a video compression standard designed as part of the MPEG-H project. ITU-T and ISO/IEC published the HEVC/H.265 standard in 2013 (version 1), 2014 (version 2), 2015 (version 3), and 2016 (version 4). Versatile Video Coding (VVC/H.266) is a video compression standard intended as a successor to HEVC. ITU-T and ISO/IEC published the VVC/H.266 standard in 2020 (version 1) and 2022 (version 2). AOMedia Video 1 (AV1) is an open video coding format designed as an alternative to HEVC. On Jan. 8, 2019, a validated version 1.0.0 with Errata 1 of the specification was released. Enhanced Compression Model (ECM) is a video coding standard that is currently under development. ECM aims to significantly improve compression efficiency beyond existing standards like HEVC/H.265 and VVC, essentially allowing for higher quality video at lower bitrates. ECM version 13 was published on Jul. 7, 2024 in MPEG 146.
The present disclosure describes, amongst other things, a set of techniques for video (image) compression related to prediction modes and transform set selection for corresponding residual blocks. Some embodiments include adaptively generating an intra predictor using a decoder side intra mode derivation (DIMD). The system (e.g., a decoder) may then check the intra predictor determined using DIMD to determine if it is within an intra mode replacement set (e.g., the set may be hardcoded, or accessed via a look-up table). If the intra predictor is within the replacement set, a final intra predictor may be derived using interpolation or matrix multiplication. Deriving the intra predictor in this manner can improve the coding accuracy (obtaining a better intra predictor) as compared to conventional intra predictor derivations (e.g., not utilizing the replacement set or performing different final intra predictor derivations). Some embodiments include using intra modes of neighboring blocks to construct an intra mode list (e.g., a most probable mode (MPM) list) for a current block. For example, if the neighboring intra mode is a position dependent intra prediction (PDP), the PDP may be added to the intra mode list (e.g., instead of or in addition to a corresponding conventional intra mode). Mapping (non-conventional) intra modes from neighboring blocks to construct an intra mode list can improve coding accuracy (e.g., by increasing diversity in the intra mode list). Some embodiments include using different sets of transform kernels (e.g., primary or secondary) based on whether DIMD is used on the current block. For example, the transform kernels may correspond to a non-separable primary or secondary transform. Using different transform kernels based on whether DIMD is applied can improve coding accuracy (e.g., selecting the most appropriate kernel based on the intra mode derivation). Some embodiments include combining an intra prediction step with a non-separable primary transform step (e.g., applying a single, combined matrix rather than 2 separate matrices). Combining the intra prediction and non-separable primary transform steps can decrease compute complexity and decrease coding time. Some embodiments include using the techniques described herein with other matrix-base coding tools (e.g., to improve the accuracy of those tools in a similar manner).
In accordance with some embodiments, a method of video decoding includes: (i) receiving a video bitstream (e.g., a coded video sequence) comprising a current block; (ii) identifying a first prediction mode for the current block (e.g., using DIMD); (iii) when the first prediction mode is a particular prediction mode, selecting a first set of transform kernels as transform kernels for the current block; (iv) when the first prediction mode is not the particular prediction mode, selecting a second set of transform kernels as the transform kernels for the current block; and (v) applying a transform for the current block using the transform kernels.
In accordance with some embodiments, a method of video decoding includes: (i) receiving a video bitstream comprising a current block; (ii) identifying a first prediction mode for the current block using a DIMD technique; (iii) determining whether the first prediction mode is in an intra mode replacement set; (iv) when the first prediction mode is in the intra mode replacement set, the intra predictor is generated using a first technique; and (v) when the first prediction mode is not in the intra mode replacement set, the intra predictor is generated using a second technique.
In accordance with some embodiments, a method of video decoding includes: (i) receiving a video bitstream comprising a current block; (ii) identifying a first prediction mode for the current block; and (iii) when the first prediction mode is a particular prediction mode, generating a transformed residual block for the current block without performing a prediction block calculation.
In accordance with some embodiments, a method of video encoding includes (i) receiving video data (e.g., a source video sequence) comprising a plurality of blocks including a current block; (ii) identifying a first prediction mode for the current block; (iii) when the first prediction mode is a particular prediction mode, selecting a first set of transform kernels as transform kernels for the current block; (iv) when the first prediction mode is not the particular prediction mode, selecting a second set of transform kernels as the transform kernels for the current block; and (v) applying a transform for the current block using the transform kernels.
In accordance with some embodiments, a computing system is provided, such as a streaming system, a server system, a personal computer system, or other electronic device. The computing system includes control circuitry and memory storing one or more sets of instructions. The one or more sets of instructions including instructions for performing any of the methods described herein. In some embodiments, the computing system includes an encoder component and a decoder component (e.g., a transcoder). In accordance with some embodiments, a non-transitory computer-readable storage medium is provided. The non-transitory computer-readable storage medium stores one or more sets of instructions for execution by a computing system. The one or more sets of instructions including instructions for performing any of the methods described herein.
Thus, devices and systems are disclosed with methods for encoding and decoding video. Such methods, devices, and systems may complement or replace conventional methods, devices, and systems for video encoding/decoding. The features and advantages described in the specification are not necessarily all-inclusive and, in particular, some additional features and advantages will be apparent to one of ordinary skill in the art in view of the drawings, specification, and claims provided in this disclosure. Moreover, it should be noted that the language used in the specification has been principally selected for readability and instructional purposes and has not necessarily been selected to delineate or circumscribe the subject matter described herein.
So that the present disclosure can be understood in greater detail, a more particular description can be had by reference to the features of various embodiments, some of which are illustrated in the appended drawings. The appended drawings, however, merely illustrate pertinent features of the present disclosure and are therefore not necessarily to be considered limiting, for the description can admit to other effective features as the person of skill in this art will appreciate upon reading this disclosure.
In accordance with common practice, the various features illustrated in the drawings are not necessarily drawn to scale, and like reference numerals can be used to denote like features throughout the specification and figures.
The present disclosure describes video/image compression techniques including selecting transform kernels based on an intra prediction mode (and/or an intra prediction mode derivation). Some embodiments include identifying a first prediction mode (and/or a prediction mode derivation) for the current block, and when the first prediction mode is a particular prediction mode, selecting a first set of transform kernels as transform kernels for the current block, whereas when the first prediction mode is not the particular prediction mode, a second set of transform kernels is selected as the transform kernels for the current block. Selecting transform kernels based on a prediction mode of the current block (and/or a prediction mode derivation technique) can improve coding accuracy (by using the prediction information to select more appropriate transform kernels).
Some embodiments include identifying a first prediction mode for the current block using a decoder-side intra mode derivation (DIMD) technique, and determining whether the first prediction mode is in an intra mode replacement set. When the first prediction mode is in the intra mode replacement set, the intra predictor is generated using a first technique, and When the first prediction mode is not in the intra mode replacement set, the intra predictor is generated using a second technique. Generating different intra predictors based on whether a prediction mode is in a mode replacement set can improve coding accuracy (by selecting more accuracy intra predictors).
Some embodiments include identifying a first prediction mode for the current block, and, when the first prediction mode is a particular prediction mode, generating a transformed residual block for the current block without performing a prediction block calculation. Generating a transformed residual block in a single step can reduce compute complexity and time (e.g., performing a single matrix multiplication operation rather than multiple sequential matrix multiplication operations).
The source device 102 includes a video source 104 (e.g., a camera component or media storage) and an encoder component 106. In some embodiments, the video source 104 is a digital camera (e.g., configured to create an uncompressed video sample stream). The encoder component 106 generates one or more encoded video bitstreams from the video stream. The video stream from the video source 104 may be high data volume as compared to the encoded video bitstream 108 generated by the encoder component 106. Because the encoded video bitstream 108 is lower data volume (less data) as compared to the video stream from the video source, the encoded video bitstream 108 requires less bandwidth to transmit and less storage space to store as compared to the video stream from the video source 104. In some embodiments, the source device 102 does not include the encoder component 106 (e.g., is configured to transmit uncompressed video to the network(s) 110).
The one or more networks 110 represents any number of networks that convey information between the source device 102, the server system 112, and/or the electronic devices 120, including for example wireline (wired) and/or wireless communication networks. The one or more networks 110 may exchange data in circuit-switched and/or packet-switched channels. Representative networks include telecommunications networks, local area networks, wide area networks and/or the Internet.
The one or more networks 110 include a server system 112 (e.g., a distributed/cloud computing system). In some embodiments, the server system 112 is, or includes, a streaming server (e.g., configured to store and/or distribute video content such as the encoded video stream from the source device 102). The server system 112 includes a coder component 114 (e.g., configured to encode and/or decode video data). In some embodiments, the coder component 114 includes an encoder component and/or a decoder component. In various embodiments, the coder component 114 is instantiated as hardware, software, or a combination thereof. In some embodiments, the coder component 114 is configured to decode the encoded video bitstream 108 and re-encode the video data using a different encoding standard and/or methodology to generate encoded video data 116. In some embodiments, the server system 112 is configured to generate multiple video formats and/or encodings from the encoded video bitstream 108. In some embodiments, the server system 112 functions as a Media-Aware Network Element (MANE). For example, the server system 112 may be configured to prune the encoded video bitstream 108 for tailoring potentially different bitstreams to one or more of the electronic devices 120. In some embodiments, a MANE is provided separate from the server system 112.
The electronic device 120-1 includes a decoder component 122 and a display 124. In some embodiments, the decoder component 122 is configured to decode the encoded video data 116 to generate an outgoing video stream that can be rendered on a display or other type of rendering device. In some embodiments, one or more of the electronic devices 120 does not include a display component (e.g., is communicatively coupled to an external display device and/or includes a media storage). In some embodiments, the electronic devices 120 are streaming clients. In some embodiments, the electronic devices 120 are configured to access the server system 112 to obtain the encoded video data 116.
The source device and/or the plurality of electronic devices 120 are sometimes referred to as “terminal devices” or “user devices.” In some embodiments, the source device 102 and/or one or more of the electronic devices 120 are instances of a server system, a personal computer, a portable device (e.g., a smartphone, tablet, or laptop), a wearable device, a video conferencing device, and/or other type of electronic device.
In example operation of the communication system 100, the source device 102 transmits the encoded video bitstream 108 to the server system 112. For example, the source device 102 may code a stream of pictures that are captured by the source device. The server system 112 receives the encoded video bitstream 108 and may decode and/or encode the encoded video bitstream 108 using the coder component 114. For example, the server system 112 may apply an encoding to the video data that is more optimal for network transmission and/or storage. The server system 112 may transmit the encoded video data 116 (e.g., one or more coded video bitstreams) to one or more of the electronic devices 120. Each electronic device 120 may decode the encoded video data 116 and optionally display the video pictures.
The encoder component 106 is configured to code and/or compress the pictures of the source video sequence into a coded video sequence 216 in real-time or under other time constraints as required by the application. In some embodiments, the encoder component 106 is configured to perform a conversion between the source video sequence and a bitstream of visual media data (e.g., a video bitstream). Enforcing appropriate coding speed is one function of a controller 204. In some embodiments, the controller 204 controls other functional units as described below and is functionally coupled to the other functional units. Parameters set by the controller 204 may include rate-control-related parameters (e.g., picture skip, quantizer, and/or lambda value of rate-distortion optimization techniques), picture size, group of pictures (GOP) layout, maximum motion vector search range, and so forth. A person of ordinary skill in the art can readily identify other functions of controller 204 as they may pertain to the encoder component 106 being optimized for a certain system design.
In some embodiments, the encoder component 106 is configured to operate in a coding loop. In a simplified example, the coding loop includes a source coder 202 (e.g., responsible for creating symbols, such as a symbol stream, based on an input picture to be coded and reference picture(s)), and a (local) decoder 210. The decoder 210 reconstructs the symbols to create the sample data in a similar manner as a (remote) decoder (when compression between symbols and coded video bitstream is lossless). The reconstructed sample stream (sample data) is input to the reference picture memory 208. As the decoding of a symbol stream leads to bit-exact results independent of decoder location (local or remote), the content in the reference picture memory 208 is also bit exact between the local encoder and remote encoder. In this way, the prediction part of an encoder interprets as reference picture samples the same sample values as a decoder would interpret when using prediction during decoding.
The operation of the decoder 210 can be the same as of a remote decoder, such as the decoder component 122, which is described in detail below in conjunction with
The decoder technology described herein, except the parsing/entropy decoding, may be to be present, in substantially identical functional form, in a corresponding encoder. For this reason, the disclosed subject matter focuses on decoder operation. Additionally, the description of encoder technologies can be abbreviated as they may be the inverse of the decoder technologies.
As part of its operation, the source coder 202 may perform motion compensated predictive coding, which codes an input frame predictively with reference to one or more previously-coded frames from the video sequence that were designated as reference frames. In this manner, the coding engine 212 codes differences between pixel blocks of an input frame and pixel blocks of reference frame(s) that may be selected as prediction reference(s) to the input frame. The controller 204 may manage coding operations of the source coder 202, including, for example, setting of parameters and subgroup parameters used for encoding the video data.
The decoder 210 decodes coded video data of frames that may be designated as reference frames, based on symbols created by the source coder 202. Operations of the coding engine 212 may advantageously be lossy processes. When the coded video data is decoded at a video decoder (not shown in
The predictor 206 may perform prediction searches for the coding engine 212. That is, for a new frame to be coded, the predictor 206 may search the reference picture memory 208 for sample data (as candidate reference pixel blocks) or certain metadata such as reference picture motion vectors, block shapes, and so on, that may serve as an appropriate prediction reference for the new pictures. The predictor 206 may operate on a sample block-by-pixel block basis to find appropriate prediction references. As determined by search results obtained by the predictor 206, an input picture may have prediction references drawn from multiple reference pictures stored in the reference picture memory 208.
Output of all aforementioned functional units may be subjected to entropy coding in the entropy coder 214. The entropy coder 214 translates the symbols as generated by the various functional units into a coded video sequence, by losslessly compressing the symbols according to technologies known to a person of ordinary skill in the art (e.g., Huffman coding, variable length coding, and/or arithmetic coding).
In some embodiments, an output of the entropy coder 214 is coupled to a transmitter. The transmitter may be configured to buffer the coded video sequence(s) as created by the entropy coder 214 to prepare them for transmission via a communication channel 218, which may be a hardware/software link to a storage device which would store the encoded video data. The transmitter may be configured to merge coded video data from the source coder 202 with other data to be transmitted, for example, coded audio data and/or ancillary data streams (sources not shown). In some embodiments, the transmitter may transmit additional data with the encoded video. The source coder 202 may include such data as part of the coded video sequence. Additional data may comprise temporal/spatial/SNR enhancement layers, other forms of redundant data such as redundant pictures and slices, Supplementary Enhancement Information (SEI) messages, Visual Usability Information (VUI) parameter set fragments, and the like.
The controller 204 may manage operation of the encoder component 106. During coding, the controller 204 may assign to each coded picture a certain coded picture type, which may affect the coding techniques that are applied to the respective picture. For example, pictures may be assigned as an Intra Picture (I picture), a Predictive Picture (P picture), or a Bi-directionally Predictive Picture (B Picture). An Intra Picture may be coded and decoded without using any other frame in the sequence as a source of prediction. Some video codecs allow for different types of Intra pictures, including, for example Independent Decoder Refresh (IDR) Pictures. A person of ordinary skill in the art is aware of those variants of I pictures and their respective applications and features, and therefore they are not repeated here. A Predictive picture may be coded and decoded using intra prediction or inter prediction using at most one motion vector and reference index to predict the sample values of each block. A Bi-directionally Predictive Picture may be coded and decoded using intra prediction or inter prediction using at most two motion vectors and reference indices to predict the sample values of each block. Similarly, multiple-predictive pictures can use more than two reference pictures and associated metadata for the reconstruction of a single block.
Source pictures commonly may be subdivided spatially into a plurality of sample blocks (for example, blocks of 4×4, 8×8, 4×8, or 16×16 samples each) and coded on a block-by-block basis. Blocks may be coded predictively with reference to other (already coded) blocks as determined by the coding assignment applied to the blocks' respective pictures. For example, blocks of I pictures may be coded non-predictively or they may be coded predictively with reference to already coded blocks of the same picture (spatial prediction or intra prediction). Pixel blocks of P pictures may be coded non-predictively, via spatial prediction or via temporal prediction with reference to one previously coded reference pictures. Blocks of B pictures may be coded non-predictively, via spatial prediction or via temporal prediction with reference to one or two previously coded reference pictures.
A video may be captured as a plurality of source pictures (video pictures) in a temporal sequence. Intra-picture prediction (often abbreviated to intra prediction) makes use of spatial correlation in a given picture, and inter-picture prediction makes uses of the (temporal or other) correlation between the pictures. In an example, a specific picture under encoding/decoding, which is referred to as a current picture, is partitioned into blocks. When a block in the current picture is similar to a reference block in a previously coded and still buffered reference picture in the video, the block in the current picture can be coded by a vector that is referred to as a motion vector. The motion vector points to the reference block in the reference picture, and can have a third dimension identifying the reference picture, in case multiple reference pictures are in use.
The encoder component 106 may perform coding operations according to a predetermined video coding technology or standard, such as any described herein. In its operation, the encoder component 106 may perform various compression operations, including predictive coding operations that exploit temporal and spatial redundancies in the input video sequence. The coded video data, therefore, may conform to a syntax specified by the video coding technology or standard being used.
In some embodiments, the decoder component 122 includes a receiver coupled to the channel 218 and configured to receive data from the channel 218 (e.g., via a wired or wireless connection). The receiver may be configured to receive one or more coded video sequences to be decoded by the decoder component 122. In some embodiments, the decoding of each coded video sequence is independent from other coded video sequences. Each coded video sequence may be received from the channel 218, which may be a hardware/software link to a storage device which stores the encoded video data. The receiver may receive the encoded video data with other data, for example, coded audio data and/or ancillary data streams, that may be forwarded to their respective using entities (not depicted). The receiver may separate the coded video sequence from the other data. In some embodiments, the receiver receives additional (redundant) data with the encoded video. The additional data may be included as part of the coded video sequence(s). The additional data may be used by the decoder component 122 to decode the data and/or to more accurately reconstruct the original video data. Additional data can be in the form of, e.g., temporal, spatial, or SNR enhancement layers, redundant slices, redundant pictures, forward error correction codes, and so on.
In accordance with some embodiments, the decoder component 122 includes a buffer memory 252, a parser 254 (also sometimes referred to as an entropy decoder), a scaler/inverse transform unit 258, an intra picture prediction unit 262, a motion compensation prediction unit 260, an aggregator 268, the loop filter unit 256, a reference picture memory 266, and a current picture memory 264. In some embodiments, the decoder component 122 is implemented as an integrated circuit, a series of integrated circuits, and/or other electronic circuitry. The decoder component 122 may be implemented at least in part in software.
The buffer memory 252 is coupled in between the channel 218 and the parser 254 (e.g., to combat network jitter). In some embodiments, the buffer memory 252 is separate from the decoder component 122. In some embodiments, a separate buffer memory is provided between the output of the channel 218 and the decoder component 122. In some embodiments, a separate buffer memory is provided outside of the decoder component 122 (e.g., to combat network jitter) in addition to the buffer memory 252 inside the decoder component 122 (e.g., which is configured to handle playout timing). When receiving data from a store/forward device of sufficient bandwidth and controllability, or from an isosynchronous network, the buffer memory 252 may not be needed, or can be small. For use on best effort packet networks such as the Internet, the buffer memory 252 may be required, can be comparatively large and/or of adaptive size, and may at least partially be implemented in an operating system or similar elements outside of the decoder component 122.
The parser 254 is configured to reconstruct symbols 270 from the coded video sequence. The symbols may include, for example, information used to manage operation of the decoder component 122, and/or information to control a rendering device such as the display 124. The control information for the rendering device(s) may be in the form of, for example, Supplementary Enhancement Information (SEI) messages or Video Usability Information (VUI) parameter set fragments (not depicted). The parser 254 parses (entropy-decodes) the coded video sequence. The coding of the coded video sequence can be in accordance with a video coding technology or standard, and can follow principles well known to a person skilled in the art, including variable length coding, Huffman coding, arithmetic coding with or without context sensitivity, and so forth. The parser 254 may extract from the coded video sequence, a set of subgroup parameters for at least one of the subgroups of pixels in the video decoder, based upon at least one parameter corresponding to the group. Subgroups can include Groups of Pictures (GOPs), pictures, tiles, slices, macroblocks, Coding Units (CUs), blocks, Transform Units (TUs), Prediction Units (PUs) and so forth. The parser 254 may also extract, from the coded video sequence, information such as transform coefficients, quantizer parameter values, motion vectors, and so forth.
Reconstruction of the symbols 270 can involve multiple different units depending on the type of the coded video picture or parts thereof (such as: inter and intra picture, inter and intra block), and other factors. Which units are involved, and how they are involved, can be controlled by the subgroup control information that was parsed from the coded video sequence by the parser 254. The flow of such subgroup control information between the parser 254 and the multiple units below is not depicted for clarity.
The decoder component 122 can be conceptually subdivided into a number of functional units, and in some implementations, these units interact closely with each other and can, at least partly, be integrated into each other. However, for clarity, the conceptual subdivision of the functional units is maintained herein.
The scaler/inverse transform unit 258 receives quantized transform coefficients as well as control information (such as which transform to use, block size, quantization factor, and/or quantization scaling matrices) as symbol(s) 270 from the parser 254. The scaler/inverse transform unit 258 can output blocks including sample values that can be input into the aggregator 268. In some cases, the output samples of the scaler/inverse transform unit 258 pertain to an intra coded block; that is: a block that is not using predictive information from previously reconstructed pictures, but can use predictive information from previously reconstructed parts of the current picture. Such predictive information can be provided by the intra picture prediction unit 262. The intra picture prediction unit 262 may generate a block of the same size and shape as the block under reconstruction, using surrounding already-reconstructed information fetched from the current (partly reconstructed) picture from the current picture memory 264. The aggregator 268 may add, on a per sample basis, the prediction information the intra picture prediction unit 262 has generated to the output sample information as provided by the scaler/inverse transform unit 258.
In other cases, the output samples of the scaler/inverse transform unit 258 pertain to an inter coded, and potentially motion-compensated, block. In such cases, the motion compensation prediction unit 260 can access the reference picture memory 266 to fetch samples used for prediction. After motion compensating the fetched samples in accordance with the symbols 270 pertaining to the block, these samples can be added by the aggregator 268 to the output of the scaler/inverse transform unit 258 (in this case called the residual samples or residual signal) so to generate output sample information. The addresses within the reference picture memory 266, from which the motion compensation prediction unit 260 fetches prediction samples, may be controlled by motion vectors. The motion vectors may be available to the motion compensation prediction unit 260 in the form of symbols 270 that can have, for example, X, Y, and reference picture components. Motion compensation may also include interpolation of sample values as fetched from the reference picture memory 266, e.g., when sub-sample exact motion vectors are in use, motion vector prediction mechanisms.
The output samples of the aggregator 268 can be subject to various loop filtering techniques in the loop filter unit 256. Video compression technologies can include in-loop filter technologies that are controlled by parameters included in the coded video bitstream and made available to the loop filter unit 256 as symbols 270 from the parser 254, but can also be responsive to meta-information obtained during the decoding of previous (in decoding order) parts of the coded picture or coded video sequence, as well as responsive to previously reconstructed and loop-filtered sample values. The output of the loop filter unit 256 can be a sample stream that can be output to a render device such as the display 124, as well as stored in the reference picture memory 266 for use in future inter-picture prediction.
Certain coded pictures, once reconstructed, can be used as reference pictures for future prediction. Once a coded picture is reconstructed and the coded picture has been identified as a reference picture (by, for example, parser 254), the current reference picture can become part of the reference picture memory 266, and a fresh current picture memory can be reallocated before commencing the reconstruction of the following coded picture.
The decoder component 122 may perform decoding operations according to a predetermined video compression technology that may be documented in a standard, such as any of the standards described herein. The coded video sequence may conform to a syntax specified by the video compression technology or standard being used, in the sense that it adheres to the syntax of the video compression technology or standard, as specified in the video compression technology document or standard and specifically in the profiles document therein. Also, for compliance with some video compression technologies or standards, the complexity of the coded video sequence may be within bounds as defined by the level of the video compression technology or standard. In some cases, levels restrict the maximum picture size, maximum frame rate, maximum reconstruction sample rate (measured in, for example megasamples per second), maximum reference picture size, and so on. Limits set by levels can, in some cases, be further restricted through Hypothetical Reference Decoder (HRD) specifications and metadata for HRD buffer management signaled in the coded video sequence.
The network interface(s) 304 may be configured to interface with one or more communication networks (e.g., wireless, wireline, and/or optical networks). The communication networks can be local, wide-area, metropolitan, vehicular and industrial, real-time, delay-tolerant, and so on. Examples of communication networks include local area networks such as Ethernet, wireless LANs, cellular networks to include GSM, 3G, 4G, 5G, LTE and the like, TV wireline or wireless wide area digital networks to include cable TV, satellite TV, and terrestrial broadcast TV, vehicular and industrial to include CANBus, and so forth. Such communication can be unidirectional, receive only (e.g., broadcast TV), unidirectional send-only (e.g., CANbus to certain CANbus devices), or bi-directional (e.g., to other computer systems using local or wide area digital networks). Such communication can include communication to one or more cloud computing networks.
The user interface 306 includes one or more output devices 308 and/or one or more input devices 310. The input device(s) 310 may include one or more of: a keyboard, a mouse, a trackpad, a touch screen, a data-glove, a joystick, a microphone, a scanner, a camera, or the like. The output device(s) 308 may include one or more of: an audio output device (e.g., a speaker), a visual output device (e.g., a display or monitor), or the like.
The memory 314 may include high-speed random-access memory (such as DRAM, SRAM, DDR RAM, and/or other random access solid-state memory devices) and/or non-volatile memory (such as one or more magnetic disk storage devices, optical disk storage devices, flash memory devices, and/or other non-volatile solid-state storage devices). The memory 314 optionally includes one or more storage devices remotely located from the control circuitry 302. The memory 314, or, alternatively, the non-volatile solid-state memory device(s) within the memory 314, includes a non-transitory computer-readable storage medium. In some embodiments, the memory 314, or the non-transitory computer-readable storage medium of the memory 314, stores the following programs, modules, instructions, and data structures, or a subset or superset thereof:
In some embodiments, the decoding module 322 includes a parsing module 324 (e.g., configured to perform the various functions described previously with respect to the parser 254), a transform module 326 (e.g., configured to perform the various functions described previously with respect to the scalar/inverse transform unit 258), a prediction module 328 (e.g., configured to perform the various functions described previously with respect to the motion compensation prediction unit 260 and/or the intra picture prediction unit 262), and a filter module 330 (e.g., configured to perform the various functions described previously with respect to the loop filter 256).
In some embodiments, the encoding module 340 includes a code module 342 (e.g., configured to perform the various functions described previously with respect to the source coder 202 and/or the coding engine 212) and a prediction module 344 (e.g., configured to perform the various functions described previously with respect to the predictor 206). In some embodiments, the decoding module 322 and/or the encoding module 340 include a subset of the modules shown in
Each of the above identified modules stored in the memory 314 corresponds to a set of instructions for performing a function described herein. The above identified modules (e.g., sets of instructions) need not be implemented as separate software programs, procedures, or modules, and thus various subsets of these modules may be combined or otherwise re-arranged in various embodiments. For example, the coding module 320 optionally does not include separate decoding and encoding modules, but rather uses a same set of modules for performing both sets of functions. In some embodiments, the memory 314 stores a subset of the modules and data structures identified above. In some embodiments, the memory 314 stores additional modules and data structures not described above.
Although
The coding processes and techniques described below may be performed at the devices and systems described above (e.g., the source device 102, the server system 112, and/or the electronic device 120). According to some embodiments, methods for using selective transform sets are described below.
As described in more detail below, if an intra prediction mode is not smooth mode, or the intra prediction mode is generating prediction samples according to a given prediction direction, the intra prediction mode may be referred to as an angular or directional mode.
As discussed above, a block may refer to a coding tree block, the largest coding block, a pre-defined fixed block size, a coding block, a prediction block, a residual block, or a transform block. An inter mode coded block (or inter block) refers to a block using a inter prediction mode or combined intra-inter prediction mode. An inter mode may also refer to a block that is coded using a block vector that is used to fetch a prediction block within the same frame, e.g., using intra block copy. An intra mode coded block (or intra block) refers to a block using an intra prediction mode or a combined intra-inter prediction mode. An intra mode list may correspond to a list of most probable intra prediction modes for a current block. Additionally, the term “partitioning” may correspond to block partitioning or transform partitioning.
As an example, a coding tree unit (CTU) may be split into coding units (CUs) by using a quad-tree structure denoted as a coding tree to adapt to various local characteristics. In some embodiments, the decision on whether to code a picture area using inter-picture (temporal) or intra-picture (spatial) prediction is made at the CU level. Each CU can be further split into one, two, or four prediction units (PUs) according to the PU splitting type. Inside a PU, the same prediction process is applied, and the relevant information may be transmitted to the decoder on a PU basis. After obtaining the residual block by applying the prediction process based on the PU splitting type, a CU can be partitioned into transform units (TUs) according to another quad-tree structure like the coding tree for the CU.
A quad-tree with nested multi-type tree using binary and ternary splits segmentation structure may be used to replace the concepts of multiple partition unit types. In the coding tree structure, a CU can have either a square or rectangular shape. A CTU is first partitioned by a quaternary tree structure. The quaternary tree leaf nodes can be further partitioned by a multi-type tree structure. An example multi-type tree structure includes four splitting types. The multi-type tree leaf nodes are called CUs, and unless the CU is too large for the maximum transform length. This means that, the CU, PU, and TU may have the same block size in the quad-tree with a nested multi-type tree coding block structure.
The coding tree scheme supports the ability for the luma and chroma to have a separate block tree structure, such as in VTM7. In some cases, for P and B slices, the luma and chroma CTBs in one CTU share the same coding tree structure. However, for I slices, the luma and chroma can have separate block tree structures. When a separate block tree mode is applied, a luma CTB is partitioned into CUs by one coding tree structure, and the chroma CTBs are partitioned into chroma CUs by another coding tree structure. This means that a CU in an I slice may include, or consist of, a coding block of the luma component or coding blocks of two chroma components, and a CU in a P or B slice may always include, or consist of, coding blocks of all three color components unless the video is monochrome.
Turning now to transforms and transform blocks, the transforms performed during decoding of the video bitstream may be inverses of the transformed performed during encoding of the video bitstream, and are sometimes referred to as “inverse transforms”. Notably, while the encoder component applies transforms, the decoder component performs the inverse transforms. Thus, in the description below, transforms described in the context of the decoder component may be the inverse of the transforms applied on the encoder side. For simplicity, the transformations described herein may be referred to as “transforms” whether performed during encoding or decoding.
Multiple transform sizes (e.g., ranging from 4-point to 64-point for each dimension) and transform shapes (e.g., square or rectangular with width/height ratio's 2:1/1:2 and 4:1/1:4) may be utilized. As described in further detail below, a transform may correspond to a primary or secondary transform and to a separable or non-separable transform. A transform set is a grouping of one or more transform types. Thus, a transform set indicates a group of multiple transform kernel/bases. Each entry in the transform set may be referred to as a transform candidate. For each block, a transform candidate selected from a transform set may be signaled or implicitly identified.
Some embodiments include methods for signaling a transform set and/or type selection for intra and/or inter coded blocks. A transform type may belong to the family of sinusoidal transforms, KLTs, or line-graph transforms (LGT). A (primary or secondary) transform may belong to the family of sinusoidal transforms (DCT's, DST's, flipped versions of DCT's and ADST's). DCT may refer to any transforms that use a transform kernel originating from the discrete cosine transform basis, and DST/ADST may refer to any transforms that use a transform kernel originating from the discrete sine transform basis.
An example primary transform may belong to the family of generalized line graph transforms (LGT) or it may be a training-based kernel. An example secondary transform set may be a grouping of one or more non-separable secondary transform kernel transform types. Unique or common secondary transform sets may be defined for each primary transform type, and/or intra or inter mode type.
Additionally, non-separable transforms can refer to primary transforms applied directly to residuals, or secondary transforms applied on the transform coefficient blocks produced by the primary transform. Transform kernels can be grouped into sets denoted by set indices and kernel indices within a set. Non-separable secondary transforms may be trained kernels applied to primary transform coefficients at the encoder or dequantized coefficients at the decoder.
A non-separable secondary transform kernel can be considered as a collection of basis vectors in a vector space. If represented as a matrix of size M×N (M rows and N columns), N corresponds to the dimension of vector space and M the number of bases. Thus, M×N can be used to represent kernel size. Examples of kernel sizes include, but are not limited to, 64×64 samples, 32×64 samples, 16×64 samples, 8×84 samples, 4×64 samples, 16×16 samples, 8×16 samples, 4×16 samples, 8×8 samples, and 4×4 samples.
A scanning order refers to the coefficient reorganization process that maps a two-dimensional primary transform coefficient array to a one-dimensional primary transform coefficient array as the input the forward secondary transform, it can also refer to the backward coefficient reorganization process that maps a one-dimensional secondary transform coefficient array back to a two-dimensional primary transform coefficient array.
An end of block (EOB) value corresponds to the position of the last significant coefficient following a given coefficient scanning order in a coded block. The EOB value may correspond to the position of the last significant (e.g., non-zero) coefficient following a given coefficient scanning order in a coded block. All coefficients in positions beyond the EOB are zero for a given coefficient scanning order. In some embodiments, if a non-separable secondary transform kernel of size M×N is applied to a coded block, the EOB value is ≤M.
A two-dimensional transform process may involve the use of hybrid transform kernels (e.g., composed of different one-dimensional transforms for each dimension of the coded residual block). Primary one-dimensional transforms may include at least one of a) 4-point, 8-point, 16-point, 32-point, 64-point discrete cosine transform; b) 4-point, 8-point, 16-point asymmetric discrete sine transforms and their flipped versions; or c) 4-point, 8-point, 16-point, 32-point identity transforms.
For a chroma component, the transform type selection may be performed in an implicit way. For intra prediction residuals, the transform type may be selected according to the intra prediction mode. For inter prediction residuals, the transform type may be selected according to the transform type selection of the co-located luma block. Therefore, for chroma component, no transform type signaling in the bitstream may be needed.
Turning now to example encoding and decoding using prediction and residual blocks,
Intra prediction explores spatial redundancy between a current block and its neighboring samples. Conventionally, intra prediction modes can be classified as directional and non-directional modes, indicating their directional or non-directional correlation between neighboring reference blocks and current block.
In a second example method, an intra prediction signal is generated using reference samples multiplied with weighted coefficients. These weighted coefficients may be trained offline and stored as a matrix.
Where F (x, y, k) is the trained coefficients and r(k) is the considered reference samples. In Equation 1, (x, y) represents the coordinates within the current prediction block, and k is the iterator going through all reference samples. The final predictor is a weighted sum of all reference samples.
In the second example method, it is worth noting that no signaling overhead is introduced. Because a predefined subset of the intra modes in the first example method is replaced with the matrix-multiplication based approach. In other words, some conventional intra modes are instituted by a matrix-multiplication based method.
In a third example method, a hybrid approach of the first interpolation-based method and the second training-based matrix multiplication approach is applied (e.g., a PDP approach). For a subset of the first intra mode in the first method, the prediction of the first method is generated by the second approach. For example, the predictor corresponding to even number intra modes in the first approach is replaced by the predictor in the second example method.
In a fourth example method, the intra prediction is generated based on the first example method. However, its intra prediction mode is not explicitly signaled. Instead, its intra prediction mode is derived from the decoder side.
The training-based methodology is not limited to the prediction stage and can also be used in a transform stage. In a fifth example method, a secondary transform is applied between forward primary transform and quantization (at encoder) and between de-quantization and inverse primary transform (at decoder side) as shown in
Multiple primary and/or secondary transform kernel sets may be trained to compact different residual information adaptively. The secondary transform set is may be determined based on the intra prediction mode in the first example method.
In a sixth example method, a single-stage non-separable primary transform is applied instead of a conventional (e.g. DCT or DST) two-stage primary transform and secondary transform to the residual block.
An example LFNST 504 that may include 16 input coefficients for a 4×4 forward LFNST or 64 input coefficients for an 8×8 forward LFNST. In an example LFNST, a 4×4 non-separable transform or an 8×8 non-separable transform is applied according to block size. For example, a 4×4 LFNST may be applied for small blocks (e.g., with a width or height that is less than 8 samples) and an 8×8 LFNST may be applied for larger blocks (e.g., with a width or height that is greater than 4 samples). An example LFNST 510 may include 8 input coefficients for a 4×4 inverse LFNST or 16 input coefficients for an 8×8 inverse LFNST.
As an example, there may be 4 transform sets and 2 non-separable transform matrices (kernels) per transform set that are used in LFNST. The mapping from the intra prediction mode to the transform set may be predefined as shown in in Table 1 below. If one of three CCLM modes (INTRA_LT_CCLM, INTRA_T_CCLM or INTRA_L_CCLM) is used for the current block (81<=IntraPredMode<=83), the transform set 0 is selected (e.g., for a current chroma block). For each transform set, the selected non-separable secondary transform candidate may be further specified by an explicitly signaled LFNST index. The index may be signaled in a bitstream once per Intra CU (e.g., after transform coefficients).
The LFNST described above may be adjusted as follows. First, more transform sets may be used for a finer granularity of directions (e.g., 35 transform sets instead of 4). Second, more transform kernels may be included in each set (e.g., 3 transform kernels instead of 2). Third, three different kernels (e.g., LFNST4, LFNST8, and LFNST16) may be defined to indicate LFNST kernel sets, which may be applied to 4×N/N×4 (N≥4), 8×N/N×8 (N≥8), and M×N (M, N≥16), respectively. Fourth, the LFNST set (LFNST_set_idx) for a given intra mode (Intra_pred_mode) may be derived according to a different formula, such as shown in Table 2.
In another example, for Intra_pred_mode<2, LFNST_set_idx is equal to 2, LFNST_set_idx=Intra_pred_mode, for Intra_pred_mode in [0,34], and LFNST_set_idx=68−Intra_pred_mode, for Intra_pred_mode in [35,66].
A non-separable primary transform (NSPT) may be useed to replace the separable DCT-II plus LFNST transform combinations, e.g., for the block shapes of 4×4, 4×8/8×4, 4×16/16×4, 8×8, 8×16/16×8, 4×32/32×4, and 8×32/32×8. The NSPT set index for a given intra mode may also be derived based on Table 2.
In some embodiments, LFNST is restricted to be applicable only if all coefficients outside the first coefficient sub-group are non-significant. In these embodiments, LFNST index coding depends on the position of the last significant coefficient. The LFNST index may be context coded (e.g., not depending on the intra prediction mode). In some embodiments, only the first bin is context coded. In some embodiments, LFNST is applied for intra CU in both intra and inter slices, and for both luma and chroma components. If a dual tree is enabled, LFNST indices for luma and chroma components may be signaled separately. For inter slice (the dual tree is disabled), a single LFNST index may be signaled and used for both luma and chroma components.
In some embodiments, when an intra subpartition (ISP) mode is selected for a current block, LFNST may be disabled and RST index may not be signaled (e.g., because performance improvement may be marginal even if RST is applied to every feasible partition block). Additionally, disabling RST for ISP-predicted residual can reduce encoding complexity. LFNST may also be disabled (and the index not signaled) when a matrix-based intra prediction (MIP) mode is selected.
A large CU (e.g., greater than 64×64) may be implicitly split (TU tiling) due to existing maximum transform size restrictions (64×64). An LFNST index search could increase data buffering by four times for a certain number of decode pipeline stages. Therefore, the maximum size allowed for LFNST may be restricted (e.g., to 64×64). In some embodiments, LFNST is enabled with DCT2 only.
In some embodiments, separable transforms are applied on intra residual and inter residual samples. In some embodiments, an intra secondary transform (IST) scheme is customized for a video coding library (e.g., used for transforming intra residual blocks). The IST scheme can efficiently capture directional patterns in intra residual samples with lower complexity compared to non-separable primary transforms. In an IST scheme the nominal intra prediction angles can be used to categorize the IST kernels.
In some embodiments, a secondary transform method (e.g., IST) is applied to the primary transform coefficient block before applying quantization at the encoder (e.g., for intra prediction residual blocks of a luma component). Accordingly, a secondary inverse transform can be applied to a dequantized transform coefficient block before applying the inverse primary transform at the decoder. In some embodiments, IST is not applied to the chroma color components. The use of IST in the encoding and decoding process is illustrated in
In some embodiments, 12 secondary transform sets (or IST sets) are defined, each containing 3 secondary transform kernels. In some embodiments, for each intra coded transform block, the nominal intra prediction mode and primary transform type may be identified, then the IST set is selected based on Table 3 below. In some embodiments, for Paeth prediction mode and recursive intra prediction modes, IST is neither applied nor signaled.
Given an IST set with 3 kernels, there are four encoder options: 1) no secondary transform, 2) secondary transform using the first transform kernel in the given IST set, 3) secondary transform kernel using the second transform kernel in the given IST set, and 4) secondary transform kernel using the third transform kernel in the given IST set. The encoder may signal the selection using the using a syntax element (e.g., ist_idx). At the decoder, the value of syntax element is parsed, and, given the IST set and value associated with ist_idx, the secondary transform kernel is identified. The syntax element (ist_idx) may be signaled for each luma transform block after the signaling of primary transform type. For example, the signaling of ist_idx may be performed if at least one of the following is true: (i) the current block is an intra coded luma transform block, (ii) the primary transform type is DCT in both dimensions or ADST in both dimensions, (iii) the intra prediction mode is neither Paeth prediction mode nor recursive intra prediction mode, (iv) the transform partitioning depth is 0, and (v) the EOB position falls within the low-frequency transform coefficient region where secondary transform is applicable. In some embodiments, the entropy coding context for ist_idx is derived based on the transform block size. Conceptually, IST may be considered as another name for LFSNT. As one of skill in the art would appreciate, in the present disclosure, IST and LFNST may be replaceable for each other.
The secondary transforms (whether referred to as LFNST, NSPT or IST) may not be limited to only apply on intra predicted blocks. When the block is an inter predicted block or predicted using intra block copy mode, a decoded side intra mode derivation (DIMD) may be applied based on a template, as shown in
For the convenience of description, the most frequently used mode M0 (e.g., derived either by collecting a histogram of a template or the occurrence of neighboring blocks) is named as the primary derived mode from decoder side. However, the derived intra mode may include more than one intra mode.
In some embodiments, the intra mode is derived from the most frequent intra mode occurring in neighboring blocks (e.g., including both adjacent and non-adjacent neighboring blocks). For the purposes of deriving the intra mode, the adjacent and non-adjacent neighboring block positions might be predefined. The DIMD modes described herein are examples and in some embodiments, other DIMD methods are employed.
The system receives (602) a video bitstream comprising a current block. The system identifies (604) a first prediction mode for the current block. When the first prediction mode is a particular prediction mode, the system selects (606) a first set of transform kernels as transform kernels for the current block. When the first prediction mode is not the particular prediction mode, the system selects (608) a second set of transform kernels as the transform kernels for the current block. The system applies (610) a transform for the current block using the transform kernels. In this way, a separate set of non-separable transform kernels may be used when the intra mode is derived from decoder side using DIMD. The transform kernels may be used for a non-separable secondary transform or a single-stage non-separable primary transform.
In some embodiments, the fusion of individual predictor (in DIMD) is applied when using the separated set of a secondary transform kernel or non-separable primary transform kernel.
In some embodiments, an intra predictor is generated adaptively based on intra mode derived from decoder side using DIMD. In some embodiments, when an intra mode, M, is derived from the decoder side, and M is in the intra mode replacement subset in PDP, the intra predictor of this derived mode is generated using a first interpolation method rather than using a replacement matrix multiplication in PDP.
In some embodiments, when an intra mode, M, is derived from the decoder side and M is in the intra mode replacement subset in PDP, the intra predictor of this derived mode M is generated using replacement matrix multiplication in PDP. As an example, the fusion in DIMD is not applied. Instead, the final predictor is the predictor corresponding to mode M0 generated based on replacement matrix multiplication using the trained coefficients. As another example, the fusion method in DIMD is kept. However, the individual predictor corresponding to M0-M4 is replaced by the replacement matrix multiplication using the trained coefficients.
In some embodiments, when an intra mode M is derived from the decoder side and M is in a replacement intra mode subset, the predictor is generated with another separate set of trained coefficients using matrix multiplication, rather than replacement matrix multiplication in PDP. As an example, the fusion in DIMD is not applied. Instead, the final predictor is the predictor corresponding to mode M0 generated based on the separate set of trained coefficients using matrix multiplication. As another example, the fusion method in DIMD is kept. However, the individual predictor corresponding to M0-M4 is replaced by the separate set of trained coefficients using matrix multiplication.
In some embodiments, when deriving intra modes from neighboring block, the derived intra mode is mapped to construct an intra mode candidate list for the current block.
In some embodiments, when the derived intra mode is from PDP, the PDP is added to the candidate list.
In some embodiments, when the derived intra mode is from PDP, the PDP is considered as an explicit intra prediction mode and is added to the candidate list.
In some embodiments, when the derived intra mode is from PDP, the PDP is considered as a specific intra mode and is added to the candidate list. As an example, the derived mode M from PDP is considered as Planar mode. As another example, the derived mode M from PDP is considered as DC.
In some embodiments, a single trained method is used instead of a sequential application of a first method (e.g., matrix multiplication, DIMD, or PDP) followed by a transform method (e.g., NSPT).
In some embodiments, an intra prediction is combined together with non-separable primary transform and therefore a transformed residual block is calculated directly from reference sample without separate prediction block calculation step. For example, this method may be applied for all intra prediction modes using matrix multiplication. In another example, this method is applied for at least a subset of all intra prediction modes using PDP.
In some embodiments, a decoder-side intra mode derivation method is used to determine the transform set for the pre-defined sub-set of intra modes in the second example method, where for these intra modes the intra prediction signal is generated using a matrix-multiplication based method.
In some embodiments, the decoder-side intra mode is derived based on the prediction signal using a matrix-multiplication based method. In some embodiments, the determination of transform set using decoder-side intra mode is applied conditionally based on other coding information. For example, it is only applied when current block size is smaller than a pre-defined value N (e.g. N is 1024 samples). In another example, it is only applied when current block's ratio between width and height is smaller than a pre-defined value R (e.g. R is 4). In another example, it is only applied when current block width and height is smaller than a pre-defined value W and H. (e.g. W=32, H=32). In another example, it is only applied when the template size (T1 and/or T2) meets a pre-defined size condition (e.g. the size T1=T2=2).
In some embodiments, the predefined subset of intra modes includes planar mode. In some embodiments, the predefined subset of intra modes includes DC mode. In some embodiments, the predefined subset of intra modes includes mode equals to (2+4*k), where k=[0,16]. In some embodiments, the predefined subset of intra modes includes mode equals to (2+2*k), where k=[0,32].
In some embodiments, the determined transform set is a primary transform set. In some embodiments, the determined transform set is a secondary transform set. In some embodiments, the determination of the transform set is based on a predefined table between the derived intra mode and the determined transform set.
In some embodiments, multiple intra modes are derived with the decoder-side intra mode derivation approach. This decoder-side intra mode derivation approach analyzes the most probably modes from the predicted signal using the matrix-multiplication approach. For example, a syntax may be signaled in the bitstream to indicate which intra mode within the derived intra modes is used to determine transform set.
In some embodiments, a separate set of transform kernel is trained for the predefined subset of intra modes in the second method, where for these intra modes the intra prediction signal is generated using a matrix-multiplication based method. In some embodiments, the predefined subset of intra modes are separated from the conventional intra modes in the first example method, and they are signaled explicitly in a bitstream.
The above methods and techniques can also be used for other training-based coding tools (e.g., other matrix-based tools).
The system receives (652) video data comprising a plurality of blocks including a current block. The system identifies (654) a first prediction mode for the current block. When the first prediction mode is a particular prediction mode, the system selects (656) a first set of transform kernels as transform kernels for the current block. When the first prediction mode is not the particular prediction mode, the system selects (658) a second set of transform kernels as the transform kernels for the current block. The system applies (660) a transform for the current block using the transform kernels. As described previously, the encoding process may mirror the decoding processes described herein (e.g., transform kernel selection). For brevity, those details are not repeated here.
Although
Turning now to some example embodiments.
In some embodiments, a multiple transform set selection method is applied for intra blocks coded with LFNST/NSPT. In this way, CUs coded with decoder-side intra mode derivation (DIMD), template-based intra mode derivation (TIMD), MIP, spatial geometric partition mode (SGPM), enhanced intra prediction (EIP), and IntraTMP modes can use an alternative LFNST/NSPT transform set. The transform set selection may be signaled in the bitstream.
In some embodiments, a coding component is configured to select, for CUs coded with DIMD, TIMD, MIP, SGPM EIP, and IntraTMP, a LFNST/NSPT transform set from 2 or more candidate sets. For example, if the current block is coded with DIMD, TIMD, MIP, SGPM, IntraTMP and LFNST/NSPT is used, an additional more bin is used to indicate whether the first or second candidate transform set is selected. In some embodiments, the first candidate transform set is the default transform set (e.g., is the same as the current ECM design). In some embodiments, the second candidate is derived by a DIMD process with neighboring reconstructed pixels. For example, if a TIMD coded block applies fusion, the second TIMD IPM may be used to derive the second candidate set. As another example, for an SGPM coded block, the two IPM that SGPM uses may be used to derive the second candidate set. Simulation data on ECM-11 software with common test conditions has shown that selecting between two transform sets for LFNST/NSPT improves coding of the luma (Y) component by 0.08%. Simulation data on ECM-14 software with common test conditions has shown that selecting between two transform sets for LFNST/NSPT improves coding of the luma (Y) component by 0.14%.
In ECM-13.0, LFNST/NSPT can be applied to inter prediction (inter-coded) blocks, where the transform kernels for intra LFNST/NSPT may be reused by inter blocks. Based on the prediction signal of the current block, a histogram of gradients (HoG) may be built in a similar way as DIMD, and the first DIMD intra prediction mode (IPM) corresponding to the highest amplitude may be used to determine the LFNST/NSPT kernel set. However, as described herein, it may be beneficial to have more than one option for selecting the kernel set. For example, for a coded block, a secondary IPM may be derived for inter LFNST/NSPT. For example, in addition to the first DIMD IPM, a second DIMD IPM (corresponding to the second highest HOG amplitude) may be used as an additional IPM candidate. Simulation data on ECM-13 software with common test conditions has shown that selecting between two transform sets for LFNST/NSPT using DIMD improves coding of the luma (Y) component and chroma component (e.g., V component) by 0.02%.
An NSPT and/or LFNST kernel set may be chosen based on the transform block size and intra mode. For example, NSPT may be used for block shapes 4×4, 4×8, 4×16, 4×32, 8×8, 8×16, 8×32 and corresponding transposed block shapes, while LFNST may be used for other block shapes. As described above, NSPT and/or LFNST may be used for various intra prediction tools, such as conventional intra prediction (e.g., planar, DC, and directional intra predictions), DIMD, TIMD, SGPM, MIP, EIP, and IntraTMP. NSPT and/or LFNST may also be used for inter-coded CUs. As described previously, multiple sets of kernels may be used for NSPT and LFNST. For example, the selection of the transform kernels may be based on whether the CU uses conventional intra prediction, inter prediction, or TIMD, DIMD, EIP, MIP, SGPM, or IntraTMP. Simulation data on ECM-14 software with common test conditions has shown that selecting between multiple transform kernels for LFNST/NSPT based on prediction mode (e.g., conventional intra, inter-coded CUs, or non-conventional intra) improves coding of the luma (Y) component by 0.17%, improves the coding of the chroma (U) component by 0.09%, and improves the coding of the chroma (V) component by 0.01%.
Although EMC is mentioned in the previous paragraphs, one of skill in the art would recognize that the methods described herein can be used in many existing codecs, such as those mentioned in the background section.
(A1) In one aspect, some embodiments include a method (e.g., the method 600) of video decoding. In some embodiments, the method is performed at a computing system (e.g., the server system 112, the source device 102, or the electronic device 120) having memory and one or more processors. In some embodiments, the method is performed at a coding module (e.g., the coding module 320). In some embodiments, the method is performed by executing instructions stored in the memory (e.g., the memory 314) of the one or more computing systems. The method includes: (i) receiving a video bitstream (e.g., a coded video sequence) comprising a current block; (ii) identifying a first prediction mode for the current block; (iii) when the first prediction mode is a particular prediction mode, selecting a first set of transform kernels as transform kernels for the current block; (iv) when the first prediction mode is not the particular prediction mode, selecting a second set of transform kernels as the transform kernels for the current block; and (v) applying a transform for the current block using the transform kernels. For example, a separate set of non-separable primary (and/or secondary) transform kernels for the transform is used when an intra mode is derived from the decoder side. In some embodiments, applying the transform comprises applying an inverse of a transform performed at an encoder. In some embodiments, in accordance with a determination that the first prediction mode is a particular prediction mode, a first set of transform kernels is selected for the current block. In some embodiments, in accordance with a determination that the first prediction mode is not the particular prediction mode, a second set of transform kernels is selected for the current block. As an example, a decoder-side intra mode derivation method may be used to determine the transform set for a predefined subset of intra modes, where, for these intra modes, the intra prediction signal is generated using a matrix-multiplication based method.
(A2) In some embodiments of A1, the particular prediction mode is identified using a decoder-side intra mode derivation (DIMD). For example, the DIMD may include the intra prediction being generated using interpolation filter applied on reference samples without the intra prediction mode being explicitly signaled. In some embodiments, the first set of transform kernels is selected in accordance with a determination that the first prediction mode is identified using the DIMD. In some embodiments, the first set of transform kernels is selected in accordance with a determination that the first prediction mode is an unconventional intra prediction mode (e.g., not a directional, DC, or planar intra mode). As an example, multiple intra modes may be derived with the decoder-side intra mode derivation approach that analyzes the most probably modes from the predicted signal using a matrix-multiplication approach. In some embodiments, a syntax is signaled in the bitstream to indicate which intra mode within the derived intra modes is used to determine transform set.
(A3) In some embodiments of A2, applying the transform comprises applying a fusion of individual predictors for the DIMD. For example, the fusion of individual predictors in the DIMD method is applied when using the first (separate) set of transform kernels.
(A4) In some embodiments of any of A1-A3, the transform is a non-separable primary transform (NSPT). For example, a single-stage non-separable primary transform is applied instead of a conventional (e.g., DCT or DST) two-stage primary transform and secondary transform to the residual block. In some embodiments, the transform is from a primary transform set.
(A5) In some embodiments of any of A1-A3, the transform is a low-frequency non-separable transform (LFNST). In some embodiments, the transform is from a secondary transform set.
(A6) In some embodiments of any of A1-A3, the transform is a secondary transform. For example, a secondary transform is applied between forward primary transform and quantization (at encoder) and between de-quantization and inverse primary transform (at decoder side).
(A7) In some embodiments of any of A1-A6, the particular prediction mode is identified using a matrix-based approach.
(A8) In some embodiments of A7, applying the transform for the current block comprises combining the matrix-based prediction with a non-separable primary transform. For example, a combined matrix is generated from the combination of the intra prediction and the non-separable primary transform. In some embodiments, the particular prediction mode and transform are applied in a single step (e.g., rather than a sequential application). in some embodiments, a matrix corresponding to the combination of the particular prediction mode and the transform is stored at a decoder component and used when the particular prediction mode and the transform are to be performed on the current block. As an example, the intra prediction is combined together with a non-separable primary transform and therefore a transformed residual block is calculated directly from the reference sample without a separate prediction block calculation step. In some embodiments, the intra prediction mode is combined with the non-separable primary transform for all intra prediction modes. In some embodiments, the intra prediction mode is combined with the non-separable primary transform for only a subset of all intra prediction modes (e.g., according to a PDP mode).
(A9) In some embodiments of any of A1-A8, the first prediction mode is identified using a position dependent prediction (PDP) approach. For example, the PDP approach may include a hybrid approach of using an interpolation filter applied on reference samples and generating intra prediction signals by using reference samples multiplied with weighted coefficients.
(A10) In some embodiments of any of A1-A9, the method further comprises: (i) deriving an intra mode for the current block based on a neighboring block of the current block; and (ii) populating an intra mode candidate list with the derived intra mode, where the first prediction mode is selected from the intra mode candidate list. For example, when deriving an intra mode from a neighboring block, the derived intra mode is mapped to construct an intra mode candidate list for the current block.
(A11) In some embodiments of A10, the intra mode is derived using a PDP approach. For example, when the intra mode is derived using a PDP approach, the derived intra mode is added to the candidate list. As an example, when the intra mode derived using a PDP approach, the derived intra mode is considered as a specific intra mode. In some embodiments, the derived intra mode is mapped to a conventional intra mode (e.g., a non-directional intra mode). For example, the derived intra mode is considered as a planar mode or a DC mode. For example, the decoder-side intra mode is derived based on the prediction signal using the matrix-multiplication based method.
(A12) In some embodiments of any of A1-A1, the method further comprises generating an intra predictor using the first prediction mode, including: (i) identifying the first prediction mode using DIMD; (ii) determining whether the first prediction mode is in an intra mode replacement set; (iii) when the first prediction mode is in the intra mode replacement set, the intra predictor is generated using a first technique; and (iv) when the first prediction mode is not in the intra mode replacement set, the intra predictor is generated using a second technique. For example, the intra predictor is generated adaptively based on intra mode derived from decoder side in the DIMD. In some embodiments, the intra mode replacement set is hardcoded at the decoder. In some embodiments, the intra mode replacement set is stored as a look-up table.
(A13) In some embodiments of A12, the first technique is an interpolation technique. For example, when an intra mode derived from the decoder side DIMD is in the intra mode replacement subset (e.g., in the PDP method), the intra predictor of this derived mode is generated using a first interpolation method (e.g., conventional interpolation) rather than using a replacement matrix multiplication in the PDP method.
(A14) In some embodiments of A12, the first technique is a replacement matrix multiplication technique. For example, when an intra mode derived from the decoder side DIMD is in the intra mode replacement subset, the intra predictor of this derived mode is generated using replacement matrix multiplication of a hybrid approach of interpolation and training-based matrix multiplication (e.g., a PDP approach). In some embodiments, the fusion of intra predictors in the DIMD approach is not applied. For example, the fusion in the DIMD is not applied. Instead, the final predictor is the predictor corresponding to mode M0 generated based on replacement matrix multiplication using the trained coefficients. In some embodiments, the fusion in the DIMD is applied to replacement predictors. For example, the individual predictors corresponding to M0-M4 of DIMD are replaced using the replacement matrix multiplication with trained coefficients.
(A15) In some embodiments of A14, the replacement matrix multiplication technique uses a set of trained coefficients corresponding to a combined DIMD and PDP approach. For example, when an intra mode derived from the decoder side DIMD approach is in the replacement intra mode subset, the predictor is generated with another separate set of trained coefficients using matrix multiplication, rather than replacement matrix multiplication of a PDP approach. In some embodiments, the fusion of intra predictors in the DIMD approach is not applied. For example, fusion in the DIMD approach is not applied. Instead, the final predictor is the predictor corresponding to mode M0 generated based on the separate set of trained coefficients using matrix multiplication. In some embodiments, the fusion method in the DIMD approach is applied. However, the individual predictor corresponding to M0-M4 are replaced by the separate set of trained coefficients using matrix multiplication.
(A16) In some embodiments of any of A1-A15, the second set of transform kernels is selected from a plurality of transform kernel sets based on coding information. For example, the determination of transform set using decoder-side intra mode is applied conditionally based on other coding information. For example, the determination of the transform set is based on a predefined table between the derived intra mode and the determined transform set. In some embodiments, the second set of transform kernels is trained using a predefined set of intra modes. For example, a separate set of transform kernel is trained for the predefined subset of intra modes, where, for these intra modes, the intra prediction signal is generated using a matrix-multiplication based method.
(A17) In some embodiments of A16, the coding information comprises at least one of a size of current block, an aspect ratio of the current block, and a template size. In one example, the second set of transform kernels is selected when current block size is smaller than a predefined value, N (e.g., N is 1024 samples). In another example, the second set of transform kernels is selected when current block's ratio between width and height is smaller than a predefined value R (e.g., R is 4). In another example, the second set of transform kernels is selected when the current block width and height are smaller than a predefined value W and H (e.g., W=32, H=32). In another example, the second set of transform kernels is selected when the template size (T1 and/or T2) meets a predefined size condition (e.g. the size T1=T2=2).
(A18) In some embodiments of any of A1-A17, the particular prediction mode is one of a subset of prediction modes. For example, the predefined subset of intra modes includes one or more of: a planar mode, a DC mode, modes equal to (2+4*k), where k=[0,16], and modes equal to (2+2*k), where k=[0,32]. In some embodiments, the predefined subset of intra modes are separated from conventional intra modes and are signaled explicitly in a bitstream.
(B1) In another aspect, some embodiments include a method (e.g., the method 650) of video encoding. In some embodiments, the method is performed at a computing system (e.g., the server system 112) having memory and one or more processors. In some embodiments, the method is performed at a coding module (e.g., the coding module 320). The method includes: (i) receiving video data (e.g., a source video sequence) comprising a plurality of blocks including a current block; (ii) identifying a first prediction mode for the current block; (iii) when the first prediction mode is a particular prediction mode, selecting a first set of transform kernels as transform kernels for the current block; (iv) when the first prediction mode is not the particular prediction mode, selecting a second set of transform kernels as the transform kernels for the current block; and (v) applying a transform for the current block using the transform kernels.
(B2) In some embodiments of B1, the particular prediction mode is identified using a decoder-side intra mode derivation (DIMD).
(B3) In some embodiments of B1 or B2, the transform is a non-separable primary transform (NSPT) or a low-frequency non-separable transform (LFNST).
(B4) In some embodiments of any of B1-B3, the particular prediction mode is identified using a matrix-based approach.
(B5) In some embodiments of any of B1-B4, the method further comprises any of A1-A18 above.
(C1) In another aspect, some embodiments include a method of processing visual media data. In some embodiments, the method is performed at a computing system (e.g., the server system 112) having memory and one or more processors. In some embodiments, the method is performed at a coding module (e.g., the coding module 320). The method includes: (i) obtaining a source video sequence that comprises a plurality of frames; and (ii) performing a conversion between the source video sequence and a video bitstream of visual media data according to a format rule, where the video bitstream comprises a plurality of blocks including a current block; and the format rule specifies that: (a) a first prediction mode for the current block is to be identified; (b) when the first prediction mode is a particular prediction mode, a first set of transform kernels is selected as transform kernels for the current block; (c) when the first prediction mode is not the particular prediction mode, a second set of transform kernels is selected as the transform kernels for the current block; and (d) a transform is applied for the current block using the transform kernels.
(D1) In another aspect, some embodiments include a method of video decoding. In some embodiments, the method is performed at a computing system (e.g., the server system 112, the source device 102, or the electronic device 120) having memory and one or more processors. In some embodiments, the method is performed at a coding module (e.g., the coding module 320). In some embodiments, the method is performed by executing instructions stored in the memory (e.g., the memory 314) of the one or more computing systems. The method includes: (i) receiving a video bitstream comprising a current block; (ii) identifying a first prediction mode for the current block using a decoder-side intra mode derivation (DIMD) technique; (iii) determining whether the first prediction mode is in an intra mode replacement set; (iv) when the first prediction mode is in the intra mode replacement set, the intra predictor is generated using a first technique; and (v) when the first prediction mode is not in the intra mode replacement set, the intra predictor is generated using a second technique. In some embodiments, in accordance with a determination that the first prediction mode is in the intra mode replacement set, the intra predictor is generated using a first technique. In some embodiments, in accordance with a determination that the first prediction mode is not in the intra mode replacement set, the intra predictor is generated using a second technique, different than the first technique.
(E1) In another aspect, some embodiments include a method of video decoding. In some embodiments, the method is performed at a computing system (e.g., the server system 112, the source device 102, or the electronic device 120) having memory and one or more processors. In some embodiments, the method is performed at a coding module (e.g., the coding module 320). In some embodiments, the method is performed by executing instructions stored in the memory (e.g., the memory 314) of the one or more computing systems. The method includes: (i) receiving a video bitstream comprising a current block; (ii) identifying a first prediction mode for the current block; and (iii) when the first prediction mode is a particular prediction mode, generating a transformed residual block for the current block without performing a prediction block calculation. For example, a single matrix is applied to the samples to generate the transformed residual block. In some embodiments, the single matrix corresponds to a combination of a prediction block matrix and a transform matrix.
In another aspect, some embodiments include a computing system (e.g., the server system 112) including control circuitry (e.g., the control circuitry 302) and memory (e.g., the memory 314) coupled to the control circuitry, the memory storing one or more sets of instructions configured to be executed by the control circuitry, the one or more sets of instructions including instructions for performing any of the methods described herein (e.g., A1-A18, B1-B5, C1, D1, and E1 above). In yet another aspect, some embodiments include a non-transitory computer-readable storage medium storing one or more sets of instructions for execution by control circuitry of a computing system, the one or more sets of instructions including instructions for performing any of the methods described herein (e.g., A1-A18, B1-B5, C1, D1, and E1 above).
Unless otherwise specified, any of the syntax elements described herein may be high-level syntax (HLS). As used herein, HLS is signaled at a level that is higher than a block level. For example, HLS may correspond to a sequence level, a frame level, a slice level, or a tile level. As another example, HLS elements may be signaled in a video parameter set (VPS), a sequence parameter set (SPS), a picture parameter set (PPS), an adaptation parameter set (APS), a slice header, a picture header, a tile header, and/or a CTU header.
It will be understood that, although the terms “first,” “second,” etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the claims. As used in the description of the embodiments and the appended claims, the singular forms “a,” “an” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will also be understood that the term “and/or” as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items. It will be further understood that the terms “comprises” and/or “comprising,” when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
As used herein, the term “when” can be construed to mean “if” or “upon” or “in response to determining” or “in accordance with a determination” or “in response to detecting” that a stated condition precedent is true, depending on the context. Similarly, the phrase “if it is determined [that a stated condition precedent is true]” or “if [a stated condition precedent is true]” or “when [a stated condition precedent is true]” can be construed to mean “upon determining” or “in response to determining” or “in accordance with a determination” or “upon detecting” or “in response to detecting” that the stated condition precedent is true, depending on the context. As used herein, N refers to a variable number. Unless explicitly stated, different instances of N may refer to the same number (e.g., the same integer value, such as the number 2) or different numbers.
The foregoing description, for purposes of explanation, has been described with reference to specific embodiments. However, the illustrative discussions above are not intended to be exhaustive or limit the claims to the precise forms disclosed. Many modifications and variations are possible in view of the above teachings. The embodiments were chosen and described in order to best explain principles of operation and practical applications, to thereby enable others skilled in the art.
This application claims priority to U.S. Provisional Patent Application No. 63/623,197, entitled “Video Coding with Training-based Coding Tool,” filed Jan. 19, 2024, and U.S. Provisional Patent Application No. 63/633,563, entitled “Method and Apparatus for Selective Transform Set based on Decoder-Side Intra Mode Derivation,” filed Apr. 12, 2024, each of which is hereby incorporated by reference in its entirety.
| Number | Date | Country | |
|---|---|---|---|
| 63623197 | Jan 2024 | US | |
| 63633563 | Apr 2024 | US |
