Complexity reduction method for neural network-based video coding tools

Abstract

A video coder is configured to perform a neural network (NN)-based filtering process on a reconstructed block of video data. In one example, a video coder may receive a picture of video data and reconstruct a block of the picture of video data to produce a reconstructed block. The video coder may perform an NN-based filtering process on the reconstructed block to generate a filtered block, wherein the NN-based filtering process includes performing a plurality of separable convolutions to approximate a multi-dimensional convolution.

Description

[0001] This application claims the benefit of U.S. Application No. 18 / 442,955, filed on February 15, 2024, and U.S. Provisional Patent Application No. 63 / 485,862, filed on February 17, 2023, the entire contents of which are incorporated herein by reference. U.S. Application No. 18 / 442,955, filed on February 15, 2024, claims the benefit of U.S. Provisional Patent Application No. 63 / 485,840, filed on February 17, 2023. Technical Field

[0002] This disclosure relates to video encoding and video decoding. Background Art

[0003] Digital video capabilities can be incorporated into a wide variety of devices, including digital televisions, digital live broadcast systems, wireless broadcast systems, personal digital assistants (PDAs), laptop or desktop computers, tablet computers, e-book readers, digital cameras, digital recording devices, digital media players, video game devices, video game consoles, cellular or satellite radio telephones (so-called "smartphones"), video teleconferencing devices, video streaming devices, and the like. Digital video devices implement video coding techniques, such as those described in the standards defined by MPEG-2, MPEG-4, ITU-T H.263, ITU-T H.264 / MPEG-4 (Part 10, Advanced Video Coding (AVC)), ITU-T H.265 / High Efficiency Video Coding (HEVC), ITU-T H.266 / Versatile Video Coding (VVC), and extensions of such standards, as well as proprietary video codecs / formats such as AOMedia Video 1 (AV1) developed by the Alliance for Open Media. By implementing such video coding techniques, video devices can more efficiently transmit, receive, encode, decode, and / or store digital video information.

[0004] Video coding techniques include spatial (intra-picture) prediction and / or temporal (inter-picture) prediction to reduce or eliminate redundancy inherent in video sequences. For block-based video coding, a video slice (e.g., a video picture or a portion of a video picture) can be divided into video blocks, which can also be referred to as coding tree units (CTUs), coding units (CUs), and / or coding nodes. Video blocks in an intra-coded (I) slice of a picture are encoded using spatial prediction relative to reference samples in neighboring blocks in the same picture. Video blocks in an inter-coded (P or B) slice of a picture can use spatial prediction relative to reference samples in neighboring blocks in the same picture, or temporal prediction relative to reference samples in other reference pictures. Pictures can be referred to as frames, and reference pictures can be referred to as reference frames. Summary of the Invention

[0005] In general, this disclosure describes techniques for video decoding. In particular, this disclosure describes methods, techniques, and apparatus that can reduce the computational complexity and memory bandwidth requirements of neural network (NN)-based video decoding tools. The example techniques described herein relate to NN-based filtering. However, the techniques of this disclosure are applicable to any NN-based video decoding tool that uses input data with certain statistical properties. In some instances, the NN-based decoding tool can be a convolutional NN (CNN)-based video decoding tool, such as a CNN-based filter. The techniques of this disclosure can be used in the context of advanced video codecs, such as extensions of VVC or next-generation video coding standards, and / or any other video codec.

[0006] According to the techniques of this disclosure, a video decoder can be configured to utilize separable convolutions instead of multidimensional convolutions. For example, two separable one-dimensional convolutions can be used to replace a 3x3 convolution in any portion of a NN-based filter. The use of separable convolutions can reduce computational complexity and memory bandwidth requirements.

[0007] In one example, a method of decoding video data includes: receiving a picture of the video data; reconstructing a block of the picture of the video data to produce a reconstructed block; and performing a NN-based filtering process on the reconstructed block to produce a filtered block, wherein the NN-based filtering process includes performing multiple separable convolutions to approximate a multidimensional convolution.

[0008] In another example, an apparatus configured to decode video data includes a memory configured to store a picture of the video data; and processing circuitry in communication with the memory, the processing circuitry configured to receive the picture of the video data, reconstruct blocks of the picture of the video data to produce reconstructed blocks, and perform an NN-based filtering process on the reconstructed blocks to produce filtered blocks, wherein the NN-based filtering process includes performing multiple separable convolutions to approximate a multidimensional convolution.

[0009] In another example, a device configured to decode video data includes: means for receiving a picture of the video data; means for reconstructing blocks of the picture of the video data to produce reconstructed blocks; and means for performing a NN-based filtering process on the reconstructed blocks to produce filtered blocks, wherein the NN-based filtering process includes performing multiple separable convolutions to approximate a multidimensional convolution.

[0010] In another example, the present disclosure describes a non-transitory computer-readable storage medium storing instructions that, when executed, cause one or more processors to: receive a picture of video data; reconstruct blocks of the picture of the video data to produce reconstructed blocks; and perform a NN-based filtering process on the reconstructed blocks to produce filtered blocks, wherein the NN-based filtering process includes performing multiple separable convolutions to approximate a multidimensional convolution.

[0011] The details of one or more examples are set forth in the accompanying drawings and the description that follows. Other features, objects, and advantages will be apparent from the description, drawings, and claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 is a block diagram illustrating an example video encoding and decoding system that may perform the techniques of this disclosure.

[0013] Figure 2 is a block diagram illustrating an example hybrid video coding framework.

[0014] Figure 3 is a conceptual diagram illustrating a hierarchical prediction structure using a group of pictures (GOP) of size 16.

[0015] Figure 4 is a block diagram showing an example convolutional neural network (CNN) based filter with four layers.

[0016] Figure 5 is a block diagram illustrating an example CNN-based filter with padded input samples and supplementary data.

[0017] Figure 6 is a block diagram illustrating another example CNN-based filter with padded input samples and supplementary data.

[0018] Figure 7 It shows Figure 6 An example of a neural network filter for attention block diagram of a residual block.

[0019] Figure 8 It shows Figure 6 Block diagram of an example of a spatial attention layer.

[0020] Figure 9 is a block diagram illustrating an example of a simplified CNN-based filter architecture with padded input samples and supplementary data using a residual block structure.

[0021] Figure 10 It shows Figure 9 Block diagram of an example residual block structure for an example of .

[0022] Figure 11is a block diagram illustrating an example of a simplified CNN-based filter architecture with padded input samples and supplementary data using a filter block structure.

[0023] Figure 12 It shows Figure 11 Block diagram of an example filter block structure for an example.

[0024] Figure 13 is a block diagram illustrating an example of multidimensional convolution decomposition.

[0025] Figure 14 is a block diagram illustrating another example of multidimensional convolution decomposition.

[0026] Figure 15 is a block diagram illustrating an example video encoder that may perform the techniques of this disclosure.

[0027] Figure 16 is a block diagram illustrating an example video decoder that may perform the techniques of this disclosure.

[0028] Figure 17 is a flowchart illustrating an example method for encoding a current block according to techniques of this disclosure.

[0029] Figure 18 is a flowchart illustrating an example method for decoding a current block according to techniques of this disclosure.

[0030] Figure 19 is a flowchart illustrating an example method for coding video data using NN-based filters, in accordance with techniques of this disclosure. DETAILED DESCRIPTION

[0031] Video encoding is typically a lossy process. For example, during the video encoding process, blocks of video data may be encoded using quantization and transforms. Typically, quantization of a value involves reducing the number of least significant bits of the value, which is typically irreversible. This bit reduction is typically performed in a manner that avoids detectable loss. However, sometimes, such loss can result in detectable artifacts in the video data, such as blocking artifacts.

[0032] Filtering can be applied to decoded and / or reconstructed video data to enhance the video data, which can improve the output video data. For example, filtering can compensate for blocking artifacts or other losses in the video data. Research has shown that filtering techniques based on neural networks (NNs) have a strong ability to improve decoded and / or reproduced video data. NN-based filtering techniques can be highly complex and require significant processing power to implement effectively.

[0033] This disclosure describes simplifications that can be applied to NN-based filtering techniques. Application of these simplifications can reduce the processing performed by one or more processors to perform NN-based filtering. In this way, the techniques of this disclosure can improve the performance of video decoding devices. Similarly, these techniques can enable more devices to perform NN-based filtering, thereby generally improving the field of video decoding.

[0034] For example, according to the techniques of this disclosure, a video decoder can be configured to utilize separable convolutions instead of multi-dimensional convolutions. For example, two separable one-dimensional convolutions can be used to replace a 3x3 convolution in any portion of a NN-based filter. The use of separable convolutions can reduce computational complexity and memory bandwidth requirements.

[0035] Figure 1 is a block diagram illustrating an example video encoding and decoding system 100 that can perform the techniques of this disclosure. In general, the techniques of this disclosure relate to transcoding (encoding and / or decoding) video data, including filtering the video data using NN-based techniques. Generally, video data includes any data used to process video. Thus, video data can include original, unencoded video, encoded video, decoded (e.g., reconstructed) video, and video metadata (such as signaling data).

[0036] like Figure 1 As shown in , in this example, system 100 includes a source device 102 that provides encoded video data to be decoded and displayed by a destination device 116. In particular, source device 102 provides the video data to destination device 116 via a computer-readable medium 110. Source device 102 and destination device 116 can be or include any of a variety of devices, such as a desktop computer, a notebook computer (i.e., a laptop computer), a mobile device, a tablet computer, a set-top box, a telephone handset such as a smartphone, a television, a camera, a display device, a digital media player, a video game console, a video streaming device, a broadcast receiver device, etc. In some cases, source device 102 and destination device 116 can be equipped for wireless communication and, therefore, can be referred to as wireless communication devices.

[0037] exist Figure 1In the example of , source device 102 includes video source 104, memory 106, video encoder 200 and output interface 108. Destination device 116 includes input interface 122, video decoder 300, memory 120 and display device 118. According to the present disclosure, the video encoder 200 of source device 102 and the video decoder 300 of destination device 116 can be configured to apply techniques for NN-based video decoding and filtering. Therefore, source device 102 represents an example of a video encoding device, while destination device 116 represents an example of a video decoding device. In other examples, the source device and destination device can include other components or arrangements. For example, source device 102 can receive video data from an external video source such as an external camera. Similarly, destination device 116 can be connected to an external display device with an interface, rather than including an integrated display device.

[0038] like Figure 1 The system 100 shown in FIG. 1 is merely an example. In general, any digital video encoding and / or decoding device can implement techniques for NN-based video decoding and filtering. Source device 102 and destination device 116 are merely examples of such decoding devices, wherein source device 102 generates decoded video data for transmission to destination device 116. This disclosure refers to a "coding" device as a device that performs decoding (e.g., encoding and / or decoding) of data. Thus, video encoder 200 and video decoder 300 represent examples of decoding devices, and in particular, represent a video encoder and a video decoder, respectively. In some examples, source device 102 and destination device 116 can operate in a substantially symmetrical manner, such that each of source device 102 and destination device 116 includes video encoding and decoding components. Thus, system 100 can support one-way or two-way video transmission between source device 102 and destination device 116, for example, for video streaming, video playback, video broadcasting, or video telephony.

[0039] Typically, video source 104 represents a source of video data (i.e., raw, unencoded video data) and provides a sequential series of pictures (also referred to as "frames") of the video data to video encoder 200, which encodes the data for the pictures. Video source 104 of source device 102 may include a video capture device, such as a camera, a video archive containing previously captured raw video, and / or a video feed interface for receiving video from a video content provider. As another alternative, video source 104 may generate computer graphics-based data as the source video, or a combination of real-time video, archived video, and computer-generated video. In each case, video encoder 200 may encode captured, pre-captured, or computer-generated video data. Video encoder 200 may rearrange the pictures from the order in which they are received (sometimes referred to as "display order") into a decoding order for decoding. Video encoder 200 may generate a bitstream comprising the encoded video data. Source device 102 may then output the encoded video data onto computer-readable medium 110 via output interface 108 for receipt and / or retrieval by, for example, input interface 122 of destination device 116 .

[0040] Memory 106 of source device 102 and memory 120 of destination device 116 represent general purpose memory. In some examples, memories 106 and 120 can store raw video data, e.g., raw video from video source 104 and raw decoded video data from video decoder 300. Additionally or alternatively, memories 106 and 120 can store software instructions executable by, for example, video encoder 200 and video decoder 300, respectively. Although memory 106 and memory 120 are shown separately from video encoder 200 and video decoder 300 in this example, it should be understood that video encoder 200 and video decoder 300 can also include internal memory for functionally similar or equivalent purposes. Furthermore, memories 106 and 120 can store, for example, encoded video data output from video encoder 200 and input to video decoder 300. In some examples, portions of memories 106 and 120 can be allocated as one or more video buffers, e.g., to store raw, decoded, and / or encoded video data.

[0041] The computer-readable medium 110 can represent any type of medium or device capable of transmitting encoded video data from the source device 102 to the destination device 116. In one example, the computer-readable medium 110 represents a communication medium that enables the source device 102 to send the encoded video data directly to the destination device 116 in real time, for example, via a radio frequency network or a computer-based network. According to a communication standard such as a wireless communication protocol, the output interface 108 can demodulate the transmission signal including the encoded video data, and the input interface 122 can demodulate the received transmission signal. The communication medium can include any wireless or wired communication medium, such as a radio frequency (RF) spectrum or one or more physical transmission lines. The communication medium can form part of a packet-based network, such as a local area network, a wide area network, or a global network such as the Internet. The communication medium can include a router, a switch, a base station, or any other device that can be used to facilitate communication from the source device 102 to the destination device 116.

[0042] In some examples, source device 102 may output the encoded data from output interface 108 to storage device 112. Similarly, destination device 116 may access the encoded data from storage device 112 via input interface 122. Storage device 112 may include any of a variety of distributed or locally accessed data storage media, such as a hard drive, Blu-ray disc, DVD, CD-ROM, flash memory, volatile or non-volatile memory, or any other suitable digital storage medium for storing encoded video data.

[0043] In some examples, source device 102 may output the encoded video data to file server 114 or another intermediate storage device, which may store the encoded video data generated by source device 102. Destination device 116 may access the stored video data from file server 114 via streaming or downloading.

[0044] The file server 114 may be any type of server device capable of storing encoded video data and transmitting the encoded video data to the destination device 116. The file server 114 may represent a web server (e.g., for a website), a server configured to provide file transfer protocol services (e.g., File Transfer Protocol (FTP) or File Delivery over Unidirectional Transport (FLUTE)), a content delivery network (CDN) device, a hypertext transfer protocol (HTTP) server, a Multimedia Broadcast Multicast Service (MBMS) or enhanced MBMS (eMBMS) server, and / or a network attached storage (NAS) device. The file server 114 may additionally or alternatively implement one or more HTTP streaming protocols, such as Dynamic Adaptive Streaming over HTTP (DASH), HTTP Live Streaming (HLS), Real Time Streaming Protocol (RTSP), HTTP Dynamic Streaming, etc.

[0045] The destination device 116 may access the encoded video data from the file server 114 through any standard data connection, including an Internet connection. This may include a wireless channel (e.g., a Wi-Fi connection), a wired connection (e.g., a digital subscriber line (DSL), a cable modem, etc.), or a combination of both, suitable for accessing the encoded video data stored on the file server 114. The input interface 122 may be configured to operate according to any one or more of the various protocols discussed above for retrieving or receiving media data from the file server 114, or other such protocols for retrieving media data.

[0046] The output interface 108 and the input interface 122 may represent wireless transmitters / receivers, modems, wired networking components (e.g., Ethernet cards), wireless communication components that operate according to any of the various IEEE 802.11 standards, or other physical components. In examples where the output interface 108 and the input interface 122 include wireless components, the output interface 108 and the input interface 122 may be configured to transmit data (such as encoded video data) according to a cellular communication standard (such as 4G, 4G-LTE (Long Term Evolution), Advanced LTE, 5G, etc.). In some examples where the output interface 108 includes a wireless transmitter, the output interface 108 and the input interface 122 may be configured to transmit data (such as encoded video data) according to other wireless standards (such as the IEEE 802.11 specifications, the IEEE 802.15 specifications (e.g., ZigBee TM ), Bluetooth TMStandards, etc.) to transmit data (such as encoded video data). In some examples, source device 102 and / or destination device 116 may include respective system-on-chip (SoC) devices. For example, source device 102 may include an SoC device for performing the functions attributed to video encoder 200 and / or output interface 108, and destination device 116 may include an SoC device for performing the functions attributed to video decoder 300 and / or input interface 122.

[0047] The techniques of the present disclosure may be applied to video decoding to support any of a variety of multimedia applications, such as over-the-air television broadcasting, cable television transmission, satellite television transmission, Internet streaming video transmission (such as Dynamic Adaptive Streaming over HTTP (DASH)), digital video encoded onto a data storage medium, decoding of digital video stored on a data storage medium, or other applications.

[0048] The input interface 122 of the destination device 116 receives an encoded video bitstream from the computer-readable medium 110 (e.g., a communication medium, a storage device 112, a file server 114, etc.). The encoded video bitstream may include signaling information defined by the video encoder 200 and also used by the video decoder 300, such as syntax elements with values ​​describing characteristics and / or processing of video blocks or other coding units (e.g., slices, pictures, groups of pictures, sequences, etc.). The display device 118 displays decoded pictures of the decoded video data to a user. The display device 118 may represent any of a variety of display devices, such as a liquid crystal display (LCD), a plasma display, an organic light emitting diode (OLED) display, or another type of display device.

[0049] Despite Figure 1 Not shown, but in some examples, the video encoder 200 and the video decoder 300 may each be integrated with an audio encoder and / or audio decoder, and may include appropriate MUX-DEMUX units or other hardware and / or software to process multiplexed streams including both audio and video in a common data stream.

[0050] The video encoder 200 and the video decoder 300 can each be implemented as any of a variety of suitable encoder and / or decoder circuits, such as one or more microprocessors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), discrete logic, software, hardware, firmware, or any combination thereof. When the technology is partially implemented in software, the device can store instructions for the software in an appropriate non-transitory computer-readable medium and use one or more processors to execute the instructions in the hardware to perform the technology of the present disclosure. Each of the video encoder 200 and the video decoder 300 can be included in one or more encoders or decoders, any of which can be integrated as part of a combined encoder / decoder (CODEC) in the corresponding device. A device including the video encoder 200 and / or the video decoder 300 can implement the video encoder 200 and / or the video decoder 300 in a processing circuit such as an integrated circuit and / or a microprocessor. Such a device can be a wireless communication device (such as a cellular phone), or any other type of device described herein.

[0051] The video encoder 200 and the video decoder 300 may operate in accordance with a video coding standard, such as the ITU-T H.265 standard (also known as the High Efficiency Video Coding (HEVC) standard) or an extension thereof, such as the multi-view and / or scalable video coding extensions. Alternatively, the video encoder 200 and the video decoder 300 may operate in accordance with other proprietary or industry standards, such as the ITU-T H.266 standard, also known as Versatile Video Coding (VVC). In other examples, the video encoder 200 and the video decoder 300 may operate in accordance with a proprietary video codec / format, such as AOMedia Video 1 (AV1), an extension of AV1, and / or a subsequent version of AV1 (e.g., AV2). In other examples, the video encoder 200 and the video decoder 300 may operate in accordance with other proprietary formats or industry standards. However, the techniques of the present disclosure are not limited to any particular coding standard or format. In general, the video encoder 200 and the video decoder 300 may be configured to perform the techniques of the present disclosure in conjunction with any video coding technique that uses a neural network.

[0052] Generally speaking, the video encoder 200 and video decoder 300 can perform block-based decoding of a picture. The term "block" generally refers to a structure that includes data to be processed (e.g., to be encoded, decoded, or otherwise used in the encoding and / or decoding process). For example, a block can include a two-dimensional matrix of luma and / or chroma data samples. Generally speaking, the video encoder 200 and video decoder 300 can decode video data represented in a YUV (e.g., Y, Cb, Cr) format. That is, rather than decoding the red, green, and blue (RGB) data for samples of a picture, the video encoder 200 and video decoder 300 can decode luma and chroma components, where the chroma components can include both red and blue hue chroma components. In some examples, the video encoder 200 converts the received RGB formatted data to a YUV representation before encoding, and the video decoder 300 converts the YUV representation to an RGB format. Alternatively, a pre-processing unit and a post-processing unit (not shown) can perform these conversions.

[0053] In general, the present disclosure may relate to decoding (e.g., encoding and decoding) of a picture to include the process of encoding or decoding the data of the picture. Similarly, the present disclosure may relate to decoding of a block of a picture to include the process of encoding or decoding the data of the block, for example, prediction and / or residual decoding. A coded video bitstream typically includes a series of values ​​for syntax elements that represent decoding decisions (e.g., decoding modes) and the partitioning of the picture into blocks. Therefore, references to decoding a picture or block should generally be understood as decoding the values ​​of the syntax elements used to form the picture or block.

[0054] HEVC defines various blocks, including coding units (CUs), prediction units (PUs), and transform units (TUs). According to HEVC, a video decoder (e.g., video encoder 200) divides a coding tree unit (CTU) into CUs according to a quadtree structure. That is, the video decoder divides the CTU and CU into four equal, non-overlapping squares, and each node of the quadtree has zero or four child nodes. A node without child nodes may be referred to as a "leaf node," and the CU of such a leaf node may include one or more PUs and / or one or more TUs. The video decoder may further partition the PUs and TUs. For example, in HEVC, the residual quadtree (RQT) represents the partitioning of the TU. In HEVC, PU represents inter-frame prediction data, and TU represents residual data. An intra-predicted CU includes intra-frame prediction information, such as an intra-frame mode indication.

[0055] As another example, the video encoder 200 and the video decoder 300 can be configured to operate according to VVC. According to VVC, a video decoder (such as the video encoder 200) divides a picture into multiple CTUs. The video encoder 200 can divide the CTU according to a tree structure (such as a quadtree-binary tree (QTBT) structure or a multi-type tree (MTT) structure). The QTBT structure removes the concept of multiple partition types, such as the separation between CU, PU and TU in HEVC. The QTBT structure includes two levels: a first level divided according to quadtree partitioning, and a second level divided according to binary tree partitioning. The root node of the QTBT structure corresponds to the CTU. The leaf nodes of the binary tree correspond to decoding units (CUs).

[0056] In the MTT partitioning structure, blocks can be partitioned using quadtree (QT) partitioning, binary tree (BT) partitioning, and one or more types of ternary tree (TT) (also known as ternary tree (TT)) partitioning. A ternary tree or ternary tree partitioning is a partitioning in which a block is divided into three sub-blocks. In some examples, the ternary tree or ternary tree partitioning divides the block into three sub-blocks without partitioning the original block via the center. The partitioning types in MTT (e.g., QT, BT, and TT) can be symmetric or asymmetric.

[0057] When operating according to the AV1 codec, the video encoder 200 and the video decoder 300 can be configured to decode video data in blocks. In AV1, the largest decoding block that can be processed is called a super block. In AV1, a super block can be 128x128 luma samples or 64x64 luma samples. However, in subsequent video decoding formats (e.g., AV2), super blocks can be defined by different (e.g., larger) luma sample sizes. In some examples, the super block is the top level of the block quadtree. The video encoder 200 can further divide the super block into smaller decoding blocks. The video encoder 200 can use square or non-square partitioning to divide the super block and other decoding blocks into smaller blocks. Non-square blocks can include N / 2xN, NxN / 2, N / 4xN, and NxN / 4 blocks. The video encoder 200 and the video decoder 300 can perform separate prediction and transform processing on each decoding block.

[0058] AV1 also defines tiles of video data. A tile is a rectangular array of superblocks that can be decoded independently of other tiles. That is, the video encoder 200 and the video decoder 300 can encode and decode coded blocks within a tile separately without using video data from other tiles. However, the video encoder 200 and the video decoder 300 can perform filtering across tile boundaries. The size of the tiles can be uniform or non-uniform. For encoder and decoder implementations, tile-based decoding can enable parallel processing and / or multi-threading.

[0059] In some examples, the video encoder 200 and the video decoder 300 may use a single QTBT or MTT structure to represent each of the luma component and the chroma components, while in other examples, the video encoder 200 and the video decoder 300 may use two or more QTBT or MTT structures, such as one QTBT / MTT structure for the luma component and another QTBT / MTT structure for the two chroma components (or two QTBT / MTT structures for the respective chroma components).

[0060] The video encoder 200 and the video decoder 300 may be configured to use quadtree partitioning, QTBT partitioning, MTT partitioning, superblock partitioning, or other partitioning structures.

[0061] In some examples, a CTU includes a coding tree block (CTB) of luma samples, two corresponding CTBs of chroma samples for a picture having three sample arrays, or a CTB of samples for a monochrome picture or a picture coded using three separate color planes, and syntax structures for coding the samples. A CTB can be an NxN block of samples for some value of N, such that the splitting of components into CTBs is a partitioning. A component is an array or a single sample from one of the three arrays (luminance and two chroma) that make up a picture in 4:2:0, 4:2:2, or 4:4:4 color format, or an array or a single sample of an array that makes up a picture in monochrome format. In some examples, a coding block is an MxN block of samples for some values ​​of M and N, such that the splitting of a CTB into a coding block is a partitioning.

[0062] Blocks (e.g., CTUs or CUs) can be grouped in various ways within a picture. As an example, a brick can refer to a rectangular area of ​​a CTU row within a particular tile in a picture. A tile can be a rectangular area of ​​a CTU within a particular tile column and a particular tile row in a picture. A tile column refers to a rectangular area of ​​a CTU with a height equal to the height of the picture and a width specified by a syntax element (e.g., such as in a picture parameter set). A tile row refers to a rectangular area of ​​a CTU with a height specified by a syntax element (e.g., such as in a picture parameter set) and a width equal to the width of the picture.

[0063] In some examples, a tile may be divided into multiple blocks, each of which may include one or more CTU rows within the tile. Tiles that are not divided into multiple bricks may also be referred to as bricks. However, bricks that are proper subsets of tiles may not be referred to as tiles. Bricks in a picture may also be arranged in slices. A slice may be an integer number of bricks of a picture that can be uniquely contained in a single Network Abstraction Layer (NAL) unit. In some examples, a slice includes several complete tiles, or a contiguous sequence of complete bricks of only one tile.

[0064] This disclosure may use "NxN" and "N by N" interchangeably to refer to the sample size of a block (such as a CU or other video block) in terms of the vertical and horizontal dimensions, for example, 16x16 samples or 16 by 16 samples. Typically, a 16x16 CU will have 16 samples in the vertical direction (y=16) and 16 samples in the horizontal direction (x=16). Likewise, an N×N CU typically has N samples in the vertical direction and N samples in the horizontal direction, where N represents a non-negative integer value. The samples in a CU may be arranged in rows and columns. Furthermore, a CU need not necessarily have the same number of samples in the horizontal direction as in the vertical direction. For example, a CU may include NxM samples, where M is not necessarily equal to N.

[0065] The video encoder 200 encodes video data representing prediction information and / or residual information and other information for a CU. The prediction information indicates how to predict the CU in order to form a prediction block for the CU. The residual information typically represents the sample-by-sample difference between the samples of the CU before encoding and the prediction block.

[0066] To predict a CU, the video encoder 200 may typically form a prediction block for the CU through inter-frame prediction or intra-frame prediction. Inter-frame prediction generally refers to predicting a CU based on data of a previously decoded picture, while intra-frame prediction generally refers to predicting a CU based on previously decoded data of the same picture. To perform inter-frame prediction, the video encoder 200 may use one or more motion vectors to generate a prediction block. The video encoder 200 may typically perform a motion search, for example, based on the difference between the CU and a reference block to identify a reference block that closely matches the CU. The video encoder 200 may calculate a difference metric using the sum of absolute differences (SAD), the sum of squared differences (SSD), the mean absolute difference (MAD), the mean squared difference (MSD), or other such difference calculations to determine whether the reference block closely matches the current CU. In some examples, the video encoder 200 may use unidirectional prediction or bidirectional prediction to predict the current CU.

[0067] Some examples of VVC also provide an affine motion compensation mode, which can be considered an inter-frame prediction mode. In the affine motion compensation mode, the video encoder 200 can determine two or more motion vectors representing non-translational motion (such as zooming in or out, rotation, perspective motion, or other irregular motion types).

[0068] To perform intra prediction, the video encoder 200 may select an intra prediction mode to generate a prediction block. Some examples of VVC provide 67 intra prediction modes, including various directional modes, as well as a planar mode and a DC mode. Typically, the video encoder 200 selects an intra prediction mode that describes neighboring samples of a current block (e.g., a block of a CU) from which samples of the current block are predicted. Assuming that the video encoder 200 decodes CTUs and CUs in raster scan order (from left to right, top to bottom), such samples may typically be above, above and to the left of the current block in the same picture as the current block.

[0069] The video encoder 200 encodes data indicating a prediction mode for the current block. For example, for inter-frame prediction mode, the video encoder 200 may encode data indicating which of various available inter-frame prediction modes to use and motion information for the corresponding mode. For unidirectional or bidirectional inter-frame prediction, for example, the video encoder 200 may encode motion vectors using Advanced Motion Vector Prediction (AMVP) or Merge Mode. The video encoder 200 may use a similar mode to encode motion vectors for affine motion compensation mode.

[0070] AV1 includes two general techniques for encoding and decoding coded blocks of video data. These two general techniques are intra prediction (e.g., intra frame prediction or spatial prediction) and inter prediction (e.g., inter frame prediction or temporal prediction). In the context of AV1, when an intra prediction coding mode is used to predict a block of a current frame of video data, the video encoder 200 and the video decoder 300 do not use video data from other frames of the video data. For most intra prediction modes, the video encoder 200 encodes the block of the current frame based on the difference between the sample values ​​in the current block and the prediction values ​​generated from reference samples in the same frame. The video encoder 200 determines the prediction values ​​generated from the reference samples based on the intra prediction coding mode.

[0071] After a prediction, such as intra-frame prediction or inter-frame prediction, of a block, the video encoder 200 may calculate residual data for the block. The residual data, such as a residual block, represents the sample-by-sample difference between the block and a prediction block for the block formed using the corresponding prediction mode. The video encoder 200 may apply one or more transforms to the residual block to produce transformed data in a transform domain rather than a sample domain. For example, the video encoder 200 may apply a discrete cosine transform (DCT), an integer transform, a wavelet transform, or a conceptually similar transform to the residual video data. In addition, the video encoder 200 may apply a secondary transform, such as a mode-dependent non-separable secondary transform (MDNSST), a signal-dependent transform, a Karhunen-Loeve transform (KLT), etc., after the first transform. The video encoder 200 generates transform coefficients after applying the one or more transforms.

[0072] As noted above, after any transform to produce transform coefficients, the video encoder 200 may perform quantization on the transform coefficients. Quantization generally refers to the process of quantizing the transform coefficients to potentially reduce the amount of data used to represent the transform coefficients, thereby providing further compression. By performing the quantization process, the video encoder 200 may reduce the bit depth associated with some or all transform coefficients. For example, the video encoder 200 may round down an n-bit value to an m-bit value during quantization, where n is greater than m. In some examples, to perform quantization, the video encoder 200 may perform a bitwise right shift on the value to be quantized.

[0073] After quantization, the video encoder 200 may scan the transform coefficients to produce a one-dimensional vector from a two-dimensional matrix including the quantized transform coefficients. The scan may be designed to place higher energy (and therefore lower frequency) transform coefficients at the front of the vector and lower energy (and therefore higher frequency) transform coefficients at the back of the vector. In some examples, the video encoder 200 may scan the quantized transform coefficients using a predefined scan order to produce a serialized vector and then entropy encode the quantized transform coefficients of the vector. In other examples, the video encoder 200 may perform adaptive scanning. After scanning the quantized transform coefficients to form a one-dimensional vector, the video encoder 200 may entropy encode the one-dimensional vector, for example, according to context-adaptive binary arithmetic coding (CABAC). The video encoder 200 may also entropy encode the values ​​of syntax elements that describe metadata associated with the encoded video data for use by the video decoder 300 when decoding the video data.

[0074] To perform CABAC, the video encoder 200 may assign context within a context model to a symbol to be transmitted. The context may relate to, for example, whether the neighboring values ​​of the symbol are zero. The probability determination may be based on the context assigned to the symbol.

[0075] The video encoder 200 may also generate syntax data, such as block-based syntax data, picture-based syntax data, and sequence-based syntax data, for the video decoder 300, for example, in a picture header, a block header, a slice header, or other syntax data, such as a sequence parameter set (SPS), a picture parameter set (PPS), or a video parameter set (VPS). The video decoder 300 may similarly decode such syntax data to determine how to decode the corresponding video data.

[0076] In this way, the video encoder 200 can generate a bitstream that includes coded video data, such as syntax elements describing the partitioning of a picture into blocks (e.g., CUs) and prediction and / or residual information for the blocks. Ultimately, the video decoder 300 can receive the bitstream and decode the coded video data.

[0077] In general, the video decoder 300 performs a reciprocal process of the process performed by the video encoder 200 to decode the encoded video data of the bitstream. For example, the video decoder 300 can use CABAC to decode the values ​​of syntax elements for the bitstream in a manner substantially similar to, but inverse to, the CABAC encoding process of the video encoder 200. The syntax elements can define partitioning information for partitioning a picture into CTUs and partitioning each CTU according to a corresponding partitioning structure (such as a QTBT structure) to define CUs of the CTU. The syntax elements can further define prediction information and residual information for a block of video data (e.g., a CU).

[0078] The residual information may be represented by, for example, quantized transform coefficients. The video decoder 300 may inverse quantize and inverse transform the quantized transform coefficients of a block to reproduce a residual block for the block. The video decoder 300 uses the signaled prediction mode (intra-frame prediction or inter-frame prediction) and associated prediction information (e.g., motion information for inter-frame prediction) to form a prediction block for the block. The video decoder 300 may then combine the prediction block and the residual block (on a sample-by-sample basis) to reproduce the original block. The video decoder 300 may perform additional processing, such as deblocking, to reduce visual artifacts along block boundaries.

[0079] The present disclosure may generally refer to "signaling" certain information (such as syntax elements). The term "signaling" may generally refer to the transmission of values ​​for syntax elements and / or other data used to decode encoded video data. That is, the video encoder 200 may signal values ​​for syntax elements in a bitstream. Generally, signaling refers to generating values ​​in a bitstream. As noted above, the source device 102 may transmit the bitstream to the destination device 116 in substantially real time or in non-real time (such as may occur when storing syntax elements to the storage device 112 for later retrieval by the destination device 116).

[0080] This disclosure describes methods, techniques, and structures that can reduce the computational complexity and / or memory bandwidth requirements of neural network (NN)-based video coding tools. The example techniques described below relate to NN-assisted loop filtering. However, the techniques of this disclosure are applicable to any NN-based video decoding tool that uses input data with certain statistical properties. The techniques of this disclosure can be used in the context of advanced video codecs, such as extensions of VVC or next-generation video decoding standards, and / or any other video codec.

[0081] In accordance with the techniques of this disclosure, video encoder 200 and video decoder 300 may be configured to perform NN-based video coding, including NN-based filtering using any combination of the techniques described below.

[0082] In-loop filter technology for video decoding

[0083] Figure 2 This is a conceptual diagram illustrating the framework of hybrid video coding. Video coding standards since H.261 are based on the so-called hybrid video coding principle, which Figure 3 The term hybrid refers to the combination of two approaches for reducing redundancy in video signals: prediction and transform coding with quantization of the prediction residuals. Prediction and transform reduce redundancy in the video signal by decorrelation, while quantization reduces the data represented by the transform coefficients by reducing their precision, ideally by removing only irrelevant details. This hybrid video coding design principle is also used in two recent standards, ITU-T H.265 / HEVC and ITU-T H.266 / VVC.

[0084] like Figure 2 As shown, a modern hybrid video coder 130 typically performs block segmentation, motion compensation or inter-picture prediction, intra-picture prediction, transform, quantization, entropy coding, and post-loop / in-loop filtering. Figure 2 In the example of , video decoder 130 includes a summation unit 134, a transform unit 136, a quantization unit 138, an entropy decoding unit 140, an inverse quantization unit 142, an inverse transform unit 144, a summation unit 146, a loop filter unit 148, a decoded picture buffer (DPB) 150, an intra-frame prediction unit 152, an inter-frame prediction unit 154, and a motion estimation unit 156.

[0085] Generally speaking, a video decoder 130 may receive input video data 132 when encoding video data. Block partitioning is used to divide a received picture (image) of video data into smaller blocks for use in prediction and transform operations. Early video coding standards used fixed block sizes, typically 16x16 samples. More recent standards, such as HEVC and VVC, employ tree-based partitioning structures to provide flexible partitioning.

[0086] Motion estimation unit 156 and inter-frame prediction unit 154 can predict input video data 132, for example, from previously decoded data from DPB 150. Motion-compensated prediction, or inter-picture prediction, exploits the redundancy that exists between pictures of a video sequence (hence the term "inter-picture"). Based on block-based motion compensation used in modern video codecs, predictions are obtained from one or more previously decoded pictures (i.e., reference pictures). The corresponding region used to generate the inter-frame prediction is indicated by motion information, including motion vectors and reference picture indices.

[0087] Summation unit 134 may calculate residual data as the difference between input video data 132 and the predicted data from intra-prediction unit 152 or inter-prediction unit 154. Summation unit 134 provides a residual block to transform unit 136, which applies one or more transforms to the residual block to produce a transform block. Quantization unit 138 quantizes the transform block to form quantized transform coefficients. Entropy coding unit 140 entropy encodes the quantized transform coefficients, along with other syntax elements (e.g., motion information or intra-prediction information), to produce an output bitstream 158.

[0088] At the same time, inverse quantization unit 142 inverse quantizes the quantized transform coefficients, and inverse transform unit 144 inverse transforms the transform coefficients to reproduce a residual block. Summation unit 146 combines the residual block with the prediction block (on a sample-by-sample basis) to produce a decoded block of video data. Loop filter unit 148 applies one or more filters (e.g., at least one of a neural network-based filter, a neural network-based loop filter, a neural network-based post-loop filter, an adaptive loop filter, or a predefined adaptive loop filter) to the decoded block to produce a filtered decoded block.

[0089] According to the techniques of this disclosure, the neural network filtering unit of loop filter unit 148 may receive data for a reconstructed picture of video data from summation unit 146 and from one or more other units of hybrid video coder 130 (e.g., transform unit 136, quantization unit 138, intra-prediction unit 152, inter-prediction unit 154, motion estimation unit 156, and / or one or more other filtering units within loop filter unit 148). For example, the neural network filtering unit may receive data from a deblocking filtering unit (also referred to as a "deblocking unit") of loop filter unit 148. The neural network filtering unit may receive, for example, a boundary strength value indicating whether a particular boundary is to be filtered for deblocking and, if so, the extent to which the boundary is to be filtered. For example, the boundary strength value may correspond to the number of samples on either side of the boundary to be modified and / or the extent to which the samples are to be modified.

[0090] In other examples, in addition to or in lieu of the boundary strength value, the neural network filter unit may receive any or all of the following: coding unit (CU) partition data, prediction unit (PU) partition data, transform unit (TU) partition data, deblocking filter data, quantization parameter (QP) data, intra-frame prediction data (e.g., reconstructed samples and / or predicted samples), inter-frame prediction data (e.g., reconstructed samples and / or predicted samples), data representing a distance between a decoded picture and one or more reference pictures, or motion information for one or more decoded blocks of a decoded picture. The deblocking filter data may further include one or more of the following: whether a long or short filter is used for deblocking, or whether a strong or weak filter is used for deblocking. The data representing the distance between a decoded picture and a reference picture may be represented as a picture order count (POC) difference between POC values ​​of the pictures.

[0091] A block of video data (e.g., a CTU or CU) may actually include multiple color components, such as a luma or "luminance" component, a blue-hued chroma or "chroma" component, and a red-hued chroma (chroma) component. The luma component may have a greater spatial resolution than the chroma components, and one of the chroma components may have a greater spatial resolution than the other chroma component. Alternatively, the luma component may have a greater spatial resolution than the chroma components, and the two chroma components may have equal spatial resolutions. For example, in a 4:2:2 format, the luma component may be twice as large horizontally as the chroma components and equal vertically to the chroma components. As another example, in a 4:2:0 format, the luma component may be twice as large horizontally and vertically as the chroma components. The various operations discussed above may generally be applied individually to each of the luma and chroma components (although certain coding information, such as motion information or intra-prediction direction, may be determined for the luma component and inherited by the corresponding chroma components).

[0092] In recent video codecs, a hierarchical prediction structure within a group of pictures (GOP) is applied to improve coding efficiency. Figure 3 An example hierarchical prediction structure for a group of pictures (GOP) size equal to 16 is shown.

[0093] Refer again Figure 2 Intra-picture prediction exploits the spatial redundancy present within a picture (hence the term "intra-picture") by deriving a prediction for a block from already coded / decoded spatially neighboring (reference) samples. Angular prediction, DC prediction, and planar or supra-plane prediction are used in recent video codecs, including AVC, HEVC, and VVC.

[0094] Hybrid video coding standards apply block transforms to prediction residuals (regardless of whether the prediction residuals come from inter-picture or intra-picture prediction). Earlier standards, including H.261, H.262, and H.263, used discrete cosine transforms (DCT). In HEVC and VVC, more transform kernels besides DCT are applied to account for different statistical information in specific video signals.

[0095] Quantization aims to reduce the precision of an input value or set of input values ​​in order to reduce the amount of data required to represent them. In hybrid video coding, quantization is typically applied to individual transformed residual samples, i.e., transform coefficients, resulting in integer coefficient levels. In recent video coding standards, the step size is derived from the so-called quantization parameter (QP), which controls fidelity and bitrate. Larger step sizes reduce bitrate but also reduce quality, resulting in, for example, blocking artifacts and blurred details in video pictures.

[0096] Entropy coding unit 140 may perform context-adaptive binary arithmetic coding (CABAC) on the encoded video. CABAC is used in recent video codecs such as AVC, HEVC, and VVC due to its high efficiency.

[0097] The filtering unit 148 may perform post-loop or in-loop filtering. Post-loop / in-loop filtering is a filtering process (or a combination of such processes) applied to a reconstructed picture to reduce coding artifacts. The input to the filtering process is typically a reconstructed picture (or a reconstructed block of a picture), which is a combination of a reconstructed residual signal (e.g., reconstructed samples), where the reconstructed samples include quantization error and prediction (e.g., prediction samples). Figure 2 As shown, the reconstructed picture after in-loop filtering is stored in a decoded picture buffer (DPB) 150 and used as a reference for inter-picture prediction of subsequent pictures.

[0098] Decoding artifacts are mainly determined by QP. Therefore, QP information is often used in the design of the filtering process. In HEVC, the in-loop filter includes deblocking filter and sample adaptive offset (SAO) filter. In VVC, the adaptive loop filter (ALF) is introduced as the third filter. The filtering process of ALF is as follows: R′(i,j)=R(i,j)+((∑ k≠0 ∑ l≠0 f(k,l)×K(R(i+k,j+l)-R(i,j),c(k,l))+64)>>7) (1) Where R(i,j) can be the sample before the filtering process, and R′(i,j) is the sample value after the filtering process. f(k,l) represents the filter coefficient, K(x,y) is the clipping function and c(k,l) represents the clipping parameter. The variables k and l are in arrive where L represents the filter length. The clipping function K(x,y)=min(y,max(-y,x)), which corresponds to the function Clip3(-y,y,x). The clipping operation introduces nonlinearity to make ALF more efficient by reducing the impact of adjacent sample values ​​that are too different from the current sample value. In VVC, the filtering parameters can be signaled in the bitstream and can be selected from a predefined filter set. The ALF filtering process can also be summarized as the following equation: R'(i,j)=R(i,j)+ALF_residual_ouput(R) (2)

[0099] Neural Network (NN)-based Filtering for Video Decoding

[0100] Figure 4 is a conceptual diagram illustrating a neural network based filter 170 having four layers. Various studies have shown that embedding a neural network (NN) into e.g. Figure 2 Compression efficiency can be improved in a hybrid video coding framework. Neural networks have been used for intra-frame and inter-frame prediction to improve prediction efficiency. In recent years, neural network-based in-loop filtering has also been a prominent research topic. In some examples, the filtering process is applied as a post-filter. In these instances, the filtering process is applied to the output picture, and the unfiltered picture can be used as a reference picture.

[0101] In addition to existing filters, the NN-based filter 170 can be applied, such as a deblocking filter, sample adaptive offset (SAO), and / or adaptive loop filter (ALF). The NN-based filter can also be applied exclusively, where the NN-based filter is designed to replace all existing filters. Additionally or alternatively, the NN-based filter (e.g., the NN-based filter 170) can be designed to supplement, enhance, or replace any or all of the other filters.

[0102] Figure 4 An example of a convolutional neural network (CNN) based filter with four layers is shown. Figure 4 The NN-based filtering process can take reconstructed samples as input, and the intermediate output is the residual samples, which are added back to the input to refine the input samples. The NN filter can use all color components (e.g., Y, U, and V, or Y, Cb, and Cr, i.e., luminance data 172A, blue chrominance 172B, and red chrominance 172C) as input 172 to exploit cross-component correlations. Different color components may share the same filter (including network structure and model parameters), or each component may have its own specific filter.

[0103] The filtering process can also be summarized as follows: R′(i,j)=R(i,j)+NN_filter_residual_ouput(R) (3) The model structure and model parameters of the NN-based filter may be predefined and stored in the video encoder 200 and the video decoder 300. The filter may also be signaled in the bitstream.

[0104] exist Figure 4 In the example of , a NN-based filter can include a series of feature extraction layers followed by output convolution. Figure 4 , the feature extraction layer may include a 3x3 convolution (conv) layer followed by a parameter rectified linear unit (PRe LU) layer. The convolution layer applies a convolution operation to the input data, which involves a filter or kernel that slides over the input data (e.g., reconstructed samples of input 172) and calculates the dot product at each location. The convolution operation essentially captures local patterns within the input data. For example, in the context of image processing, these patterns can be edges, textures, or other visual features. A filter or kernel is a small matrix of weights that is updated during the training process. By sliding this filter over the input data (or feature map from a previous layer) and calculating the dot product at each location, the convolution layer creates a feature map that encodes the spatial hierarchy and patterns detected in the input.

[0105] The output of a convolutional layer is a set of feature maps, each of which corresponds to a filter and captures a different aspect of the input data. This layer helps the neural network learn increasingly complex and abstract features as the data passes through the deeper layers of the network. Figure 4 The first 3x3 in the nomenclature 3x3conv 3x3x6x8 indicates that the convolutional layer has a 3x3 filter size (e.g., a 3x3 matrix). 3x3x6x8 refers to the input and output dimensions of the convolutional layer. 6 is the number of input channels, and 8 is the number of output channels.

[0106] The PReLU layer is an activation function used in neural networks and was introduced as a variant of the ReLU (rectified linear unit) activation function. As mentioned above, the convolutional layer outputs feature maps, each of which corresponds to a filter representing a feature detected in the input. After the convolutional layer, the PReLU layer applies the PReLU activation function to each element of the feature map produced by the convolutional layer. For positive values, the PReLU layer acts as a standard ReLU, letting the value pass through. For negative values, the PReLU layer allows small linear negative outputs instead of setting them to zero (for example, as ReLU does). This keeps neurons active and maintains gradient flow, which can be beneficial for learning in deep networks.

[0107] In summary, when a convolutional layer is followed by a PReLU layer, the convolutional layer first extracts features from the input data through a set of learned filters. The resulting feature map is then passed through the PReLU activation function, which introduces nonlinearity and helps avoid the problem of neuron death by allowing small gradients when the input is negative. This combination is effective at learning complex patterns in the data while maintaining robust gradient flow, which is especially beneficial in deeper network architectures.

[0108] Processing Unit

[0109] When applying NN-based filtering in video decoding, the entire video signal (pixel data) can be split into multiple processing units (e.g., 2D blocks), and each processing unit can be processed separately or combined with other information associated with this pixel block. Possible choices for processing units include frames, slices / tiles, CTUs, or any predefined or signaled shape and size. Typically, NN-based filtering is performed on reconstructed blocks of video data. Here, reconstructed blocks and samples can refer to decoded blocks produced by the video decoder 300 and blocks reconstructed in the reconstruction loop of the video encoder 200.

[0110] Type of input data

[0111] To further improve the performance of NN-based filtering, different types of input data can be jointly processed to produce filtered output. The input data may include (but is not limited to) reconstructed pixels / samples, predicted pixels / samples, pixels / samples after the loop filter, partition structure information, deblocking parameters (e.g., boundary strength (BS)), QP values, slice or picture type, or filter applicability or decoding mode mapping. The input data can be provided at different granularities. Luma reconstruction and prediction samples can be provided at the original resolution, while chroma samples can be provided at a lower resolution, such as for a 4:2:0 representation, or can be upsampled to luma resolution to achieve per-pixel representation. Similarly, QP, BS, partition or decoding mode information can be provided at a lower resolution, including the case where there is a single value per frame, slice or processing block (e.g., QP). In other examples, QP, BS, partition or decoding mode information can be expanded (e.g., replicated) to achieve per-pixel / sample representation.

[0112] Figure 5 An example of an architecture utilizing supplementary data is shown in . Figure 5is a block diagram illustrating an example CNN-based filter with padded input samples and supplementary data. The CNN-based filter 171 uses the pixels / samples of the processing block combined with the supplementary data as input 174. The input 174 may include four sub-blocks of interleaved luma samples (Yx4) 174A, and associated blue hue chroma (U) data 174B and red hue chroma (V) data 174C. The supplementary data includes a quantization parameter (QP) step 176 and a boundary strength (BS) 178. The area of ​​the input pixels / samples can be expanded from each side with four padded pixels / samples. The resulting size of the processing capacity is (4+64+4)x(4+64+4)x(4Y+2UV+1QP+3BS).

[0113] Relative to Figure 4 NN-based filters in

[17] , NN-based filter 171 may include two or more hidden layers that utilize both 1x1 convolutions and leaky ReLU layers. A leaky ReLU layer. Similar to a PReLU layer, a leaky ReLU layer allows a small non-zero gradient to be output when the layer is inactive. Instead of outputting zero for negative inputs, the leaky ReLU multiplies the inputs by a small constant. This small slope ensures that even inactive neurons still contribute a small amount to the network's learning, reducing the likelihood of the dead ReLU problem.

[0114] NN-based filtering with multimodal design

[0115] In order to further improve the performance of NN-based filtering, a multi-mode solution can be designed. For example, for each processing unit, the video encoder 200 can select from a set of modes based on rate-distortion optimization, and the selection can be signaled in the bitstream. Different modes can include different NN models, different values ​​of input information used as NN models, etc. In one example, the video encoder 200 and the video decoder 300 can use a NN-based filtering solution with multiple modes based on a single NN model by using different QP values ​​as input to the NN models of different modes.

[0116] Examples of NN Architectures

[0117] In one example, as described above, a NN-based filtering solution with multiple modes can be used. The structure of the network is as follows Figure 6 shown. Figure 6The NN-based filter includes a first portion comprising input 3x3 convolution filters 510A-510E and corresponding parameter rectified linear unit (PReLU) filters 512A-512E for each input to generate feature maps (e.g., the feature extraction portion of the NN filter). A connection unit 514 connects the feature maps and provides them to a fusion block 516 and a transition block 522. Figure 6 The NN-based filters in FIG5 further include: a set 528 of attention residual (AttRes) blocks 530A-530N; and a last portion (e.g., a tail portion) including a 3x3 convolution filter 550, a PReLU filter 552, a 3x3 convolution filter 554, and a pixel shuffling unit 556. The AttRes block may also be referred to as a backbone block.

[0118] In the first part (e.g., feature extraction part), different inputs are received, including quantization parameter (QP) 500, partition information (part) 502, boundary strength (BS) 504, predicted samples (pred) 506, and reconstructed samples (rec) 508. The corresponding 3x3 convolution filters 510A-510E and PReLU filters 512A-512E convolve and activate the corresponding inputs to produce feature maps. The connection unit 514 then connects the feature maps. The fusion block 516 including the 1x1 convolution filter 518 and the PReLU filter 520 fuses the connected feature maps. The transition block including the 3x3 convolution filter 524 and the PReLU filter 526 subsamples the fused input to create the output 188. The output 188 is then fed through a set 528 of attention residual blocks 530A-530N, which can include various numbers of attention residual blocks, for example 8. About Figure 7 Further explanation of the attention block. The output 189 from the last attention residual block in the set of attention residual blocks 184 is fed to the last part of the NN-based filter. In the last part, the 3x3 convolution filter 550, the PReLU filter 552, the 3x3 convolution filter 554 and the pixel shuffling unit 556 process the output 189, and the addition unit 558 combines the result with the original reconstructed sample input 508. This ultimately forms a filtered output for presentation and storage as a reference for subsequent inter-frame prediction, for example, in a decoded picture buffer (DPB). In some examples, Figure 6 The NN-based filter uses 96 feature maps.

[0119] Figure 7 It shows Figure 6 The conceptual diagram of the attention residual block. That is, Figure 7 The attention residual block 530 is depicted, which may include Figure 6Components similar to those of the attention residual blocks 530A-530N of FIG. In this example, the attention residual block 530 includes a first 3x3 convolution filter 532, a parameter rectified linear unit (PReLU) filter 534, a second 3x3 convolution filter 536, an attention block 538, and an addition unit 540. The addition unit 540 combines the output of the attention block 538 with the output 188 originally received by the convolution filter 532 to generate an output 189.

[0120] Figure 8 It shows Figure 7 Conceptual diagram of an example spatial attention layer. Figure 8 As shown, the spatial attention layer of the attention residual block 530 includes a 3x3 convolution filter 706, a PReLU filter 708, a 3x3 convolution filter 710, a size expansion unit 712, a 3x3 convolution filter 720, a PReLU filter 722, and a 3x3 convolution filter 724. The 3x3 convolution filter 706 receives the input 702, corresponding to Figure 6 Quantization parameter (QP) 500, partition information (part) 502, boundary strength (BS) 504, prediction information (pred) 506 and reconstructed block (rec) 508. The 3x3 convolution filter 720 receives Z K 704. The outputs of the size expansion unit 712 and the 3x3 convolution filter 724 are combined and then combined with the R value 730 to generate the S value 732. The S value 732 is then combined with the Z K The values ​​704 are combined to produce the output Z K+1 Value 734.

[0121] In other examples, alternative designs of NN architectures can be used. For example, Figure 6 A larger number of low-complexity residual blocks in the filter backbone of , as well as a reduced number of channels (feature maps) and removal of the attention module. This alternative convolutional neural network filter structure (e.g., for luminance filtering) is shown in Figure 9 middle. Figure 9 is a block diagram illustrating an example of a simplified CNN-based filter architecture with padded input samples and supplementary data.

[0122] Figure 9The NN-based filters include 3x3 convolution filters 810A-810E and PReLU filters 812A-812E, which convolve the corresponding inputs, namely QP 800, part 802, BS 804, Pred 806, and Rec 808, to generate a feature map (e.g., feature extraction part). The connection unit 814 connects the convolved inputs (e.g., feature maps). The fusion block 816 then uses a 1x1 convolution filter 818 and a PReLU filter 820 to fuse the connected feature maps. The transition block 822 then uses a 3x3 convolution filter 824 and a PReLU filter 826 to process the fused data.

[0123] In this example, the NN-based filter includes a set 828 of residual blocks 830A-830N (also referred to as backbone blocks), each of which can be based on Figure 10 The residual block structure 830 is constructed as follows. The residual blocks 830A-830N can be replaced by Figure 6 AttRes blocks 530A-530N. Figure 9 The examples may be used for luma filtering, but similar modifications may be made for chroma filtering as discussed below.

[0124] Figure 9 The number of residual blocks and channels included in the set 828 of can be configured differently. That is, N can be set to different values ​​and the number of channels in the residual block structure 830 can be set to a number other than 160 to achieve different performance-complexity tradeoffs. These modifications can be used to perform chroma filtering for processing of chroma channels.

[0125] The set 828 of residual blocks 830A-830N has N instances of the residual block structure 830. In one example, N can be equal to 32, so that there are 32 residual block structures. The residual blocks 830A-830N can use 64 feature maps, which is relatively Figure 6 The 96 feature maps used in the example are reduced.

[0126] Figure 10 It is an explanation Figure 9830 . In this example, the residual block structure 830 includes a first 1x1 convolution filter 832, which can increase the number of input channels to 160 before the activation layer (PReLU filter 834) processes the input channels. Therefore, the PReLU filter 834 can reduce the number of channels to 64 through this process. The second 1x1 convolution filter 836 then processes the reduced channels, followed by a 3x3 convolution filter 838. Finally, the combination unit 840 can combine the output of the 3x3 convolution filter 838 with the original input received by the residual block structure 830.

[0127] In yet another NN architecture, the residual block can be replaced by a filter block (also called a backbone block), such as Figure 11 In this example, the bypass branches around all layers in each residual block are removed, as shown in Figure 12 The number of channels and the number of filter blocks are configurable, e.g., 64 channels, 24 filter blocks, with 160 channels before and after activation.

[0128] Figure 11 To illustrate alternatives to the technology of the present disclosure Figure 6 Conceptual diagram of another example filter block structure of a set of focused residual blocks. Figure 11 The NN-based filters include 3x3 convolution filters 1010A-1010E and PReLU filters 1012A-1012E, which convolve the corresponding inputs, namely QP 1000, part 1002, BS 1004, Pred 1006, and Rec 1008, to form a feature map (e.g., feature extraction part). The connection unit 1014 connects the feature maps. The fusion block 1016 then uses a 1x1 convolution filter 1018 and a PReLU filter 1020 to fuse the connected inputs. The transition block 1022 then uses a 3x3 convolution filter 1024 and a PReLU filter 1026 to process the fused data.

[0129] In this example, the NN-based filtering unit includes N filter blocks 1030A-1030N (also referred to as backbone blocks), each of which may have the following Figure 12 The filter block structure 1030 may be substantially similar to the residual block structure 830, except that the combination unit 840 is omitted from the filter block structure 1030, so that the input is not combined with the output. Instead, the output of each residual block structure may be fed directly to the subsequent block.

[0130] The number of channels and the number of filter blocks can be configurable. In one example, it can be set to 64 channels and 32 filter blocks. As described above, the number of channels increased in each filter block 198 can be 160.

[0131] Figure 12 It is an explanation Figure 11 1030 。 In this example, the residual block structure 1030 includes a first 1x1 convolution filter 1032, which can increase the number of input channels to 160 before the activation layer (PReLU filter 1034) processes the input channels. Therefore, the PReLU filter 1034 can reduce the number of channels to 64 through this process. The second 1x1 convolution filter 1036 then processes the reduced channels, followed by the 3x3 convolution filter 1038. As discussed above, Figure 10 Compared to the residual block structure 830, the filter block structure 1030 does not include a combining unit.

[0132] Convolutions with 3x3 kernels are popular in NN-based filters. In the above architecture, 3x3xNxM convolutions are utilized in multiple segments and blocks, where the 3x3 kernel slides in the spatial (2D) domain. However, multidimensional convolutions (such as 2D kernel convolutions) introduce significant complexity. According to the technology of the present disclosure, the video encoder 200 and the video decoder 300 can be configured to utilize separable convolutions instead of multidimensional convolutions (e.g., 3x3xNxM convolutions). For example, two separable one-dimensional convolutions can be used to replace the 3x3 convolutions in any part of the NN-based filter. The use of separable convolutions can reduce computational complexity and memory bandwidth requirements.

[0133] Example

[0134] In order to avoid excessive computation and reduce the parameter set derived from multidimensional convolution, such as the 3x3 convolution in the above or similar CNN architecture (or the 2D convolution component of a higher dimensional kernel), the present disclosure describes a technique in which the video encoder 200 and the video decoder 300 are configured to utilize separable convolutions (e.g., 1D separable convolutions) generated by low-complexity approximations rather than multidimensional (e.g., 2D) convolutions that slide in the spatial direction. Although the techniques of the present disclosure are described with reference to 3x3 convolutions (e.g., 4x4, 5x5, or larger), the decomposition techniques of the present disclosure can be used for multidimensional convolutions of any size. Typically, multidimensional convolutions have a kernel size of n1 x n2 in the spatial dimension, where n1 and n2 are positive integers. The values ​​of n1 and n2 can be the same or different. Multidimensional convolutions can also have a size of K in the depth dimension (e.g., n1xn2xK). In addition, using the number of output channels M, multidimensional convolutions can be represented as a 4-D tensor of n1xn2xKxM.

[0135] Multidimensional convolution solution:

[0136] In one example of the present disclosure, a low-rank convolution approximation decomposes a 3x3xMxN convolution into a pixel-by-pixel convolution (1x1xMxR), two separable convolutions (3x1xRxR, 1x3xRxR), and another pixel-by-pixel convolution (1x1xRxN). Here, R is the rank of the approximation and can be used to adjust the performance / complexity of the approximation. The value of R can be an integer. In some examples, R can be derived as a function (ratio) of M or N or max(M,N). In some examples, R can be set equal to A*max(M,N), where A is less than 1 (e.g., 0.2, 0.5, 0.8), A is greater than 1 (e.g., 1.0, 1.2), or other values.

[0137] In a general example, a multidimensional convolution can be approximated by multiple separable convolutions by performing a first convolution of size n1x1 and performing a second convolution of size 1xn2 on the output of the first convolution.

[0138] Figure 13 An example of using separable convolution to approximate a multidimensional convolution in a backbone block of a NN-based filter is shown. In this example, the multidimensional convolution is a 2D 3x3 convolution. However, the technology of the present disclosure can be extended to convolutions of other dimensions.

[0139] Figure 13 Show where the backbone blocks come from Figure 9 and Figure 10 Example of residual block 830A. Figure 13 An example is shown in which the 3x3xKxK convolution 838 is decomposed into a series of 1D and separable convolutions in the residual block 1300. That is, the video encoder 200 and the video decoder 300 can be configured to perform the residual block 1300, which includes performing multiple separable convolutions to approximate a multi-dimensional convolution. Figure 13 In the example of , the multiple separable convolutions include a 3x1xRxR separable convolution (sep.conv) 1304 and a 1x3xRxR separable convolution 1306. In other examples, the order of the separable convolution 1304 and the separable convolution 1306 can be switched.

[0140] Therefore, in one example of the present disclosure, in order to perform multiple separable convolutions to approximate multidimensional convolution, the video encoder 200 and the video decoder can be configured to perform a first 1x1 convolution (e.g., 1x1xKxR convolution 1302), perform a first separable convolution of multiple separable convolutions (e.g., 3x1xRxR separable convolution 1304) on the output of the first 1x2 convolution, perform a second separable convolution of multiple separable convolutions (e.g., 1x3xRxR separable convolution 1306) on the output of the first separable convolution, and perform a second 1x1 convolution (e.g., 1x1xRxK convolution 1308) on the output of the second separable convolution. The number of output channels of the first 1x1 convolution 1302 is used to control the complexity approximation of the multidimensional convolution. The number of output channels of the separable convolutions 1304 and 1306 can be selected to control the complexity approximation of the multidimensional convolution. The separable convolutions 1304 and 1306 perform depth-wise convolution operations.

[0141] In another example, to perform multiple separable convolutions to approximate multi-dimensional convolution, the video encoder 200 and the video decoder 300 may receive an input, perform a 1x1xKxM convolution on the input, perform a PReLU layer on the output of the 1x1xKxM convolution, perform a 1x1xMxK convolution on the output of the PReLU layer; perform a 3x1xKxR separable convolution on the output of the 1x1xMxK convolution, perform a 1x3xRxK separable convolution on the output of the 3x1xKxR separable convolution, and perform a 1x1xRxK convolution on the output of the 1x3xRxR separable convolution.

[0142] In another example, Figure 13 An example is shown in which the 3x3xKxK convolution 838 is decomposed into a series of 1D and separable convolutions in the residual block 1300, but in which the first 1D convolution is fused with another 1D convolution. That is, the video encoder 200 and the video decoder 300 can be configured to perform a residual block 1310, which includes performing a 1x1xMxR convolution 1320, which is a fusion of a 1x1xMxK convolution (e.g., convolution 836) and a 1x1xKxR convolution (e.g., convolution 1302). In this way, the decomposition of multidimensional convolutions can be further simplified. The fusion of 1x1 convolutions in the residual block 1310 can be used for any situation in which a decomposed convolution (e.g., 1x1 fusion) is applied before or after another pixel-by-pixel convolution without any nonlinear or residual connection in between. The values ​​of M and R can be selected to control the complexity and accuracy of the approximation.

[0143] In other examples of the present disclosure, the NN-based filtering process includes cascaded (e.g., sequential) application of a backbone block. For example, a backbone block can be applied to multiple different color components. In other examples, the NN-based filtering process includes cascaded application of a backbone block applied in two or more parallel processing branches.

[0144] In one or more examples of the present disclosure, performing multiple separable convolutions to approximate a multidimensional convolution in a backbone block of a NN-based filter process includes applying an element-wise activation process as part of the multidimensional convolution. Examples of the element-wise activation process may include ReLU and PReLU functions. The PReLU function is an example of a parametrically controlled element-wise activation process.

[0145] Different algorithms for determining separable kernels can be used to determine the decomposition in place of 2D or other multidimensional kernels. In some examples, the Cancoamp / Parafac (CP) tensor decomposition can be used. Examples of other decompositions suitable for use in the present disclosure can be found in V. Lebedev, Y. Ganin, M. Rakhuba, I. Oseledets, V. Lempitsky, Accelerating Convolutional Neural Networks Using Fine-tuned CP Decomposition, ICLR 2015 V.

[0146] Alternative implementations and architectures:

[0147] In some examples, 2D convolutions of different dimensions (e.g., ZxY) or higher-dimensional convolution components can be used and replaced with corresponding separable convolutions 1xZ and Yx1.

[0148] Figure 14 is a block diagram showing another example of multi-dimensional convolution decomposition. Figure 14 As shown, the 3x3xKxK convolution 838 of the residual block 830A is approximated by the 3x1xKxR convolution 1400 and the 1x3xRxK convolution 1410. In other examples, the positions of the convolution 1400 and the convolution can be switched. R is the canonical rank of the decomposition. Lower rank means greater complexity reduction. Figure 13 Compared to the example of , the video encoder 200 and the video decoder 300 can approximate 3x3 convolutions with two separable convolutions and without performing leading or trailing 1x1 convolutions.

[0149] Integration into CNN architecture:

[0150] Although about Figure 9 and Figure 10 The residual block in describes Figure 13 , but the techniques of this disclosure can be used with any multi-dimensional convolution found anywhere in NN-based filters. For example, Figure 6-7(focus on residual blocks 530A-N) and Figure 11-12 The 3x3 convolutions in the backbone blocks (filter blocks 1030A-N) can also be approximated using multiple separable convolutions. In addition, the feature extraction portion of any NN-based filter described above (e.g., Figure 11 Convolutional layers 1010A-E and PReLU layers 1012A-E), fusion blocks (e.g., Figure 11 1016), transition blocks (e.g., Figure 11 1022), backbone blocks (e.g., Figure 11 1028) or the tail portion (e.g., Figure 11 Any multidimensional convolution (e.g., 3x3 convolution) shown in FIG1050 , PReLU 1052 , convolution 1054 , and pixel shuffling 1056 ) can be approximated using multiple separable convolutions, as described above with reference to FIG1051 . Figure 13 or Figure 14 In general, the feature extraction portion of a neural network can include any layer configured to extract features from input data. The tail portion of a neural network can include any number of layers at the end of the neural network before the output.

[0151] Breakdown training:

[0152] In some examples, the parameters of the separable convolution can be extracted from the trained parameters of the 2D convolution. An example extraction model can be implemented as follows.

[0153] For CP decomposition, the 3-dimensional (e.g., 3x3xk) convolution kernel K can be decomposed into Where R is the rank of the decomposition, and k1, k2, k3 are kernels of a certain dimension. The decomposition can be optimized by using nonlinear optimization methods (such as Gauss-Newton). The squared error is used to calculate the kernel parameters.

[0154] Alternatively, there are other methods inspired by stochastic gradient descent and machine learning practices. Consider multiple approximate 3x3 convolutions within a neural network model. These approximate convolutions can be optimized individually (e.g., sequentially trained to match the output feature map of the original 3x3 convolution and any accumulated approximation error from the previous approximate convolution in the network) or jointly optimized across the entire network (e.g., the approximate convolution is replaced into the network, and then the entire network or only all the approximate convolutions are optimized end-to-end).

[0155] This paper targets CNN architectures (e.g., similar to Figure 13 and Figure 14The 3x3 convolutional approximation technique described in

[15] can be applied to different types of architectures and modules that employ 2D convolutions in sliding windows across spatial dimensions and receive supplementary information as input data. Applications of the techniques described herein can be used not only in the backbone of a NN architecture, but also in the head blocks (e.g., feature extraction, fusion blocks, and / or transition blocks) or tail portions of the architecture.

[0156] Figure 15 is a block diagram illustrating an example video encoder 200 that may perform the techniques of this disclosure. Figure 15 This is provided for purposes of explanation and should not be considered limiting of the techniques as broadly illustrated and described in this disclosure. For purposes of explanation, this disclosure describes the video encoder 200 in terms of techniques for VVC and HEVC. However, the techniques of this disclosure may be performed by video encoding devices configured for other video coding standards and video coding formats, such as AV1 and subsequent versions of the AV1 video coding format.

[0157] exist Figure 15 In the example of FIG, the video encoder 200 includes a video data memory 230, a mode selection unit 202, a residual generation unit 204, a transform processing unit 206, a quantization unit 208, an inverse quantization unit 210, an inverse transform processing unit 212, a reconstruction unit 214, a filter unit 216, a decoded picture buffer (DPB) 218, and an entropy coding unit 220. Any or all of the video data memory 230, the mode selection unit 202, the residual generation unit 204, the transform processing unit 206, the quantization unit 208, the inverse quantization unit 210, the inverse transform processing unit 212, the reconstruction unit 214, the filter unit 216, the DPB 218, and the entropy coding unit 220 can be implemented in one or more processors or in processing circuitry. For example, the various units of the video encoder 200 can be implemented as one or more circuits or logic elements as part of a hardware circuit, or as part of a processor, ASIC, or FPGA. In addition, the video encoder 200 may include additional or alternative processors or processing circuitry to perform these and other functions.

[0158] The video data memory 230 may store video data to be encoded by the components of the video encoder 200. The video encoder 200 may receive video data from, for example, the video source 104 ( Figure 1) receives video data stored in the video data memory 230. The DPB 218 can act as a reference picture memory that stores reference video data for use when the video encoder 200 predicts subsequent video data. The video data memory 230 and the DPB 218 can be formed by any of a variety of memory devices, such as dynamic random access memory (DRAM) (including synchronous DRAM (SDRAM)), magnetoresistive RAM (MRAM), resistive RAM (RRAM), or other types of memory devices. The video data memory 230 and the DPB 218 can be provided by the same storage device or a separate storage device. In various examples, the video data memory 230 can be on-chip with the other components of the video encoder 200 (as shown), or off-chip relative to those components.

[0159] In this disclosure, references to the video data memory 230 should not be construed as limited to memory internal to the video encoder 200 (unless explicitly described as such) or memory external to the video encoder 200 (unless explicitly described as such). Rather, references to the video data memory 230 should be understood as reference memory that stores video data received by the video encoder 200 for encoding (e.g., video data for a current block to be encoded). Figure 1 The memory 106 may also provide temporary storage for the outputs of the various units of the video encoder 200 .

[0160] Show Figure 15 The various units of the video encoder 200 are described to help understand the operations performed by the video encoder 200. These units can be implemented as fixed-function circuits, programmable circuits, or a combination thereof. Fixed-function circuits refer to circuits that provide specific functions and are pre-set with respect to the operations that can be performed. Programmable circuits refer to circuits that can be programmed to perform various tasks and provide flexible functionality in the operations that can be performed. For example, a programmable circuit can execute software or firmware that causes the programmable circuit to operate in a manner defined by the instructions of the software or firmware. Fixed-function circuits can execute software instructions (e.g., to receive parameters or output parameters), but the type of operations performed by the fixed-function circuits is generally immutable. In some examples, one or more of the units can be different circuit blocks (fixed-function or programmable), and in some examples, one or more of the units can be integrated circuits.

[0161] The video encoder 200 may include an arithmetic logic unit (ALU), an elementary function unit (EFU), a digital circuit, an analog circuit, and / or a programmable core formed by a programmable circuit. In an example where the operation of the video encoder 200 is performed using software executed by a programmable circuit, the memory 106 ( Figure 1) may store instructions (eg, object code) for software that the video encoder 200 receives and executes, or another memory (not shown) within the video encoder 200 may store such instructions.

[0162] The video data memory 230 is configured to store received video data. The video encoder 200 may extract a picture of video data from the video data memory 230 and provide the video data to the residual generation unit 204 and the mode selection unit 202. The video data in the video data memory 230 may be original video data to be encoded.

[0163] The mode selection unit 202 includes a motion estimation unit 222, a motion compensation unit 224, and an intra prediction unit 226. The mode selection unit 202 may include additional functional units to perform video prediction according to other prediction modes. For example, the mode selection unit 202 may include a palette unit, an intra block copy unit (which may be part of the motion estimation unit 222 and / or the motion compensation unit 224), an affine unit, a linear model (LM) unit, etc.

[0164] The mode selection unit 202 typically coordinates multiple encoding passes to test combinations of encoding parameters and the resulting rate-distortion values ​​for such combinations. The encoding parameters may include: the division of CTUs into CUs, the prediction mode used for a CU, the transform type used for the residual data of a CU, the quantization parameter used for the residual data of a CU, etc. The mode selection unit 202 may ultimately select a encoding parameter combination that has a better rate-distortion value than other tested combinations.

[0165] The video encoder 200 may divide a picture retrieved from the video data memory 230 into a series of CTUs and encapsulate one or more CTUs into a slice. The mode selection unit 202 may divide the CTUs of the picture according to a tree structure (such as the MTT structure, QTBT structure, super block structure, or quadtree structure described above). As described above, the video encoder 200 may form one or more CUs by dividing the CTUs according to the tree structure. Such CUs are also commonly referred to as "video blocks" or "blocks."

[0166] Typically, mode select unit 202 also controls its components (e.g., motion estimation unit 222, motion compensation unit 224, and intra prediction unit 226) to generate a prediction block for the current block (e.g., the current CU, or, in HEVC, the overlapping portions of a PU and TU). To perform inter-prediction on the current block, motion estimation unit 222 may perform a motion search to identify one or more closely matching reference blocks in one or more reference pictures (e.g., one or more previously coded pictures stored in DPB 218). Specifically, motion estimation unit 222 may calculate values ​​representing how similar potential reference blocks are to the current block, such as based on sum of absolute differences (SAD), sum of squared differences (SSD), mean absolute difference (MAD), mean squared difference (MSD), and the like. Motion estimation unit 222 may typically perform these calculations using the sample-by-sample differences between the current block and the reference block under consideration. Motion estimation unit 222 may identify the reference block with the lowest value resulting from these calculations, indicating the reference block that most closely matches the current block.

[0167] Motion estimation unit 222 may form one or more motion vectors (MVs) that define the position of a reference block in a reference picture relative to the position of a current block in the current picture. Motion estimation unit 222 may then provide the motion vectors to motion compensation unit 224. For example, for unidirectional inter prediction, motion estimation unit 222 may provide a single motion vector, while for bidirectional inter prediction, motion estimation unit 222 may provide two motion vectors. Motion compensation unit 224 may then use the motion vectors to generate a prediction block. For example, motion compensation unit 224 may use the motion vectors to retrieve data for the reference block. As another example, if the motion vectors have fractional sample precision, motion compensation unit 224 may interpolate the values ​​of the prediction block based on one or more interpolation filters. Furthermore, for bidirectional inter prediction, motion compensation unit 224 may retrieve data for the two reference blocks identified by the respective motion vectors and combine the retrieved data, for example, by sample-by-sample averaging or weighted averaging.

[0168] When operating according to the AV1 video coding format, the motion estimation unit 222 and the motion compensation unit 224 may be configured to encode coding blocks of the video data (e.g., both luma coding blocks and chroma coding blocks) using translational motion compensation, affine motion compensation, overlapped block motion compensation (OBMC), and / or composite inter-frame intra-frame prediction.

[0169] As another example, for intra prediction or intra prediction decoding, the intra prediction unit 226 can generate a prediction block based on samples adjacent to the current block. For example, for directional mode, the intra prediction unit 226 can generally mathematically combine the values ​​of adjacent samples and pad these calculated values ​​across the current block in a defined direction to produce a prediction block. As another example, for DC mode, the intra prediction unit 226 can calculate the average of adjacent samples of the current block and generate a prediction block to include the obtained average for each sample of the prediction block.

[0170] When operating according to the AV1 video coding format, the intra prediction unit 226 can be configured to encode coding blocks of video data (e.g., both luma and chroma coding blocks) using directional intra prediction, non-directional intra prediction, recursive filter intra prediction, chroma prediction based on luma (CFL), intra block copy (IBC), and / or palette mode. The mode selection unit 202 may include additional functional units to perform video prediction according to other prediction modes.

[0171] Mode selection unit 202 provides the prediction block to residual generation unit 204. Residual generation unit 204 receives the original, unencoded version of the current block from video data memory 230 and the prediction block from mode selection unit 202. Residual generation unit 204 calculates the sample-by-sample difference between the current block and the prediction block. The resulting sample-by-sample difference defines a residual block for the current block. In some examples, residual generation unit 204 may also determine the difference between sample values ​​in the residual block to generate the residual block using residual differential pulse coded modulation (RDPCM). In some examples, residual generation unit 204 may be formed using one or more subtractor circuits that perform binary subtraction.

[0172] In the example where the mode select unit 202 partitions the CU into PUs, each PU may be associated with a luma prediction unit and a corresponding chroma prediction unit. The video encoder 200 and the video decoder 300 may support PUs of various sizes. As indicated above, the size of a CU may refer to the size of the luma coding block of the CU, while the size of a PU may refer to the size of the luma prediction unit of the PU. Assuming a particular CU size of 2Nx2N, the video encoder 200 may support PU sizes of 2Nx2N or NxN for intra prediction, and symmetrical PU sizes of 2Nx2N, 2NxN, Nx2N, NxN, or similar sizes for inter prediction. The video encoder 200 and the video decoder 300 may also support asymmetric partitioning for PU sizes of 2NxnU, 2NxnD, nLx2N, and nRx2N for inter prediction.

[0173] In an example where mode select unit 202 does not further split a CU into PUs, each CU may be associated with a luma coding block and a corresponding chroma coding block. As described above, the size of a CU may refer to the size of the luma coding block of the CU. Video encoder 200 and video decoder 300 may support CU sizes of 2Nx2N, 2NxN, or Nx2N.

[0174] For other video coding techniques (such as intra-block copy mode coding, affine mode coding, and linear model (LM) mode coding, to name a few examples), mode selection unit 202 generates a prediction block for the current block being encoded via the corresponding unit associated with the coding technique. In some examples, such as palette mode coding, mode selection unit 202 may not generate a prediction block, but instead generate syntax elements indicating how to reconstruct the block based on the selected palette. In such a mode, mode selection unit 202 may provide these syntax elements to entropy coding unit 220 for encoding.

[0175] As described above, the residual generation unit 204 receives video data of a current block and a corresponding prediction block. The residual generation unit 204 then generates a residual block for the current block. To generate the residual block, the residual generation unit 204 calculates the sample-by-sample difference between the prediction block and the current block.

[0176] The transform processing unit 206 applies one or more transforms to the residual block to generate a block of transform coefficients (referred to herein as a "transform coefficient block"). The transform processing unit 206 may apply various transforms to the residual block to form the transform coefficient block. For example, the transform processing unit 206 may apply a discrete cosine transform (DCT), a directional transform, a Karhunen-Loeve transform (KLT), or a conceptually similar transform to the residual block. In some examples, the transform processing unit 206 may perform multiple transforms on the residual block, for example, a primary transform and a secondary transform (such as a rotation transform). In some examples, the transform processing unit 206 does not apply a transform to the residual block.

[0177] When operating in accordance with AV1, the transform processing unit 206 may apply one or more transforms to the residual block to generate a block of transform coefficients (referred to herein as a "transform coefficient block"). The transform processing unit 206 may apply various transforms to the residual block to form the transform coefficient block. For example, the transform processing unit 206 may apply a combination of horizontal / vertical transforms that may include a discrete cosine transform (DCT), an asymmetric discrete sine transform (ADST), a flipped ADST (e.g., ADST in reverse order), and an identity transform (IDTX). When the identity transform is used, the transform is skipped in one of the vertical or horizontal directions. In some examples, the transform processing may be skipped.

[0178] Quantization unit 208 may quantize the transform coefficients in the transform coefficient block to produce a quantized transform coefficient block. Quantization unit 208 may quantize the transform coefficients of the transform coefficient block according to a quantization parameter (QP) value associated with the current block. Video encoder 200 (e.g., via mode select unit 202) may adjust the degree of quantization applied to the transform coefficient block associated with the current block by adjusting the QP value associated with the CU. Quantization may introduce a loss of information, and therefore, the quantized transform coefficients may have a lower precision than the original transform coefficients produced by transform processing unit 206.

[0179] The inverse quantization unit 210 and the inverse transform processing unit 212 may apply inverse quantization and inverse transform, respectively, to the quantized transform coefficient block to reconstruct a residual block from the transform coefficient block. The reconstruction unit 214 may generate a reconstructed block corresponding to the current block (although possibly with some degree of distortion) based on the reconstructed residual block and the prediction block generated by the mode selection unit 202. For example, the reconstruction unit 214 may add samples of the reconstructed residual block to corresponding samples of the prediction block generated by the mode selection unit 202 to generate a reconstructed block.

[0180] Filter unit 216 may perform one or more filter operations on the reconstructed block. For example, filter unit 216 may perform a deblocking operation to reduce blocking artifacts along the edges of the CU. In some examples, the operations of filter unit 216 may be skipped. Filter unit 216 may be configured to perform any of the NN-based video coding techniques described above. For example, filter unit 216 may be configured to receive a reconstructed block of a picture of video data and perform an NN-based filtering process on the reconstructed block to produce a filtered block. The NN-based filtering process includes performing multiple separable convolutions to approximate a multi-dimensional convolution.

[0181] When operating in accordance with AV1, the filter unit 216 may perform one or more filter operations on the reconstructed blocks. For example, the filter unit 216 may perform a deblocking operation to reduce blocking artifacts along the edges of the CU. In other examples, the filter unit 216 may apply a constrained directional enhancement filter (CDEF) (which may be applied after deblocking) and may include applying a non-separable, non-linear, low-pass directional filter based on the estimated edge direction. The filter unit 216 may also include a loop restoration filter applied after the CDEF and may include a separable symmetric normalized Wiener filter or a dual self-guided filter.

[0182] The video encoder 200 stores the reconstructed block in the DPB 218. For example, in an example where the operation of the filter unit 216 is not performed, the reconstruction unit 214 can store the reconstructed block in the DPB 218. In an example where the operation of the filter unit 216 is performed, the filter unit 216 can store the filtered reconstructed block in the DPB 218. The motion estimation unit 222 and the motion compensation unit 224 can retrieve a reference picture formed based on the reconstructed (and potentially filtered) block from the DPB 218 to perform inter-frame prediction on blocks of subsequently encoded pictures. In addition, the intra-frame prediction unit 226 can use the reconstructed block of the current picture in the DPB 218 to perform intra-frame prediction on other blocks in the current picture.

[0183] In general, entropy coding unit 220 may entropy encode syntax elements received from other functional components of video encoder 200. For example, entropy coding unit 220 may entropy encode quantized transform coefficient blocks from quantization unit 208. As another example, entropy coding unit 220 may entropy encode prediction syntax elements (e.g., motion information for inter-frame prediction or intra-frame mode information for intra-frame prediction) from mode selection unit 202. Entropy coding unit 220 may perform one or more entropy encoding operations on the syntax elements (which is another example of video data) to generate entropy-encoded data. For example, entropy coding unit 220 may perform a context-adaptive variable length coding (CAVLC) operation, a CABAC operation, a variable-to-variable (V2V) length coding operation, a syntax-based context-adaptive binary arithmetic coding (SBAC) operation, a probability interval partitioning entropy (PIPE) coding operation, an exponential Golomb coding operation, or another type of entropy coding operation on the data. In some examples, entropy coding unit 220 may operate in a bypass mode, in which the syntax elements are not entropy encoded.

[0184] The video encoder 200 may output a bitstream including entropy-encoded syntax elements required for reconstructing a block of a slice or picture. In particular, the entropy encoding unit 220 may output a bitstream.

[0185] According to AV1, the entropy coding unit 220 can be configured as a symbol-to-symbol adaptive multi-symbol arithmetic decoder. Syntax elements in AV1 include an alphabet of N elements, and the context (e.g., a probability model) includes a set of N probabilities. The entropy coding unit 220 can store the probabilities as an n-bit (e.g., 15-bit) cumulative distribution function (CDF). The entropy coding unit 220 can perform recursive scaling using an update factor based on the alphabet size to update the context.

[0186] The above operations are described with respect to blocks. Such descriptions should be understood as operations for luma coding blocks and / or chroma coding blocks. As described above, in some examples, the luma coding blocks and chroma coding blocks are the luma components and chroma components of a CU. In some examples, the luma coding blocks and chroma coding blocks are the luma components and chroma components of a PU.

[0187] In some examples, the operations performed for luma coding blocks do not need to be repeated for chroma coding blocks. As an example, the operations for identifying a motion vector (MV) and reference picture for a luma coding block do not need to be repeated in order to identify the MV and reference picture for a chroma block. Specifically, the MV for a luma coding block can be scaled to determine the MV for a chroma block, and the reference picture can be the same. As another example, the intra prediction process can be the same for luma coding blocks and chroma coding blocks.

[0188] Video encoder 200 represents an example of a device configured to encode video data, comprising: a memory configured to store the video data, and one or more processing units implemented in circuitry and configured to perform any combination of the techniques described above, including NN-based video decoding including NN-based filtering.

[0189] Figure 16 is a block diagram illustrating an example video decoder 300 that may perform the techniques of this disclosure. Figure 16 This is provided for purposes of explanation and is not intended to limit the techniques as broadly illustrated and described in this disclosure. For purposes of explanation, this disclosure describes the video decoder 300 in terms of techniques for VVC and HEVC. However, the techniques of this disclosure may be performed by video coding devices configured for other video coding standards.

[0190] exist Figure 16 In the example of FIG, the video decoder 300 includes: a coded picture buffer (CPB) memory 320, an entropy decoding unit 302, a prediction processing unit 304, an inverse quantization unit 306, an inverse transform processing unit 308, a reconstruction unit 310, a filter unit 312, and a DPB 314. Any or all of the CPB memory 320, the entropy decoding unit 302, the prediction processing unit 304, the inverse quantization unit 306, the inverse transform processing unit 308, the reconstruction unit 310, the filter unit 312, and the DPB 314 can be implemented in one or more processors or processing circuits. For example, the various units of the video decoder 300 can be implemented as one or more circuits or logic elements as part of a hardware circuit or as part of a processor, ASIC, or FPGA. Moreover, the video decoder 300 can include additional or alternative processors or processing circuits to perform these functions and other functions.

[0191] The prediction processing unit 304 includes a motion compensation unit 316 and an intra prediction unit 318. The prediction processing unit 304 may include additional units for performing prediction according to other prediction modes. As an example, the prediction processing unit 304 may include: a palette unit, an intra block copy unit (which may form part of the motion compensation unit 316), an affine unit, a linear model (LM) unit, etc. In other examples, the video decoder 300 may include more, fewer, or different functional components.

[0192] When operating in accordance with AV1, the motion compensation unit 316 may be configured to decode coding blocks of video data (e.g., both luma coding blocks and chroma coding blocks) using translational motion compensation, affine motion compensation, OBMC, and / or composite intra-frame inter prediction, as described above. The intra-frame prediction unit 318 may be configured to decode coding blocks of video data (e.g., both luma coding blocks and chroma coding blocks) using directional intra-frame prediction, non-directional intra-frame prediction, recursive filter intra-frame prediction, CFL, IBC, and / or palette mode, as described above.

[0193] The CPB memory 320 may store video data, such as an encoded video bitstream, to be decoded by components of the video decoder 300. For example, the video data stored in the CPB memory 320 may be obtained from the computer-readable medium 110 ( Figure 1 ). The CPB memory 320 may include a CPB that stores coded video data (e.g., syntax elements) from the coded video bitstream. In addition, the CPB memory 320 may store video data other than syntax elements for decoded pictures, such as temporary data representing outputs from various units of the video decoder 300. The DPB 314 typically stores decoded pictures, which the video decoder 300 may output and / or use as reference video data when decoding subsequent data or pictures of the coded video bitstream. The CPB memory 320 and the DPB 314 may be formed from any of a variety of memory devices, such as DRAM (including SDRAM), MRAM, RRAM, or other types of memory devices. The CPB memory 320 and the DPB 314 may be provided by the same memory device or by separate memory devices. In various examples, the CPB memory 320 may be on-chip with other components of the video decoder 300, or off-chip relative to those components.

[0194] Additionally or alternatively, in some examples, video decoder 300 may retrieve the video from memory 120 ( Figure 1) to retrieve the decoded video data. That is, memory 120 may store data along with CPB memory 320 as described above. Similarly, when some or all of the functionality of video decoder 300 is implemented in software to be executed by processing circuitry of video decoder 300, memory 120 may store instructions to be executed by video decoder 300.

[0195] Show Figure 16 The various units shown in FIG. 3 are provided to aid in understanding the operations performed by the video decoder 300. These units may be implemented as fixed function circuits, programmable circuits, or a combination thereof. Figure 15 , fixed-function circuits refer to circuits that provide specific functions and are pre-set with respect to the operations that can be performed. Programmable circuits refer to circuits that can be programmed to perform various tasks and provide flexible functionality in the operations that can be performed. For example, a programmable circuit can execute software or firmware that causes the programmable circuit to operate in a manner defined by the instructions of the software or firmware. Fixed-function circuits can execute software instructions (for example, to receive parameters or output parameters), but the type of operations performed by the fixed-function circuits is generally immutable. In some examples, one or more of the units may be different circuit blocks (fixed-function or programmable), and in some examples, one or more of the units may be integrated circuits.

[0196] The video decoder 300 may include an ALU, an EFU, digital circuits, analog circuits, and / or a programmable core formed by programmable circuits. In an example where the operation of the video decoder 300 is performed by software executed on the programmable circuits, on-chip or off-chip memory may store instructions (e.g., object code) of the software that the video decoder 300 receives and executes.

[0197] The entropy decoding unit 302 may receive encoded video data from the CPB and entropy decode the video data to reproduce syntax elements. The prediction processing unit 304, the inverse quantization unit 306, the inverse transform processing unit 308, the reconstruction unit 310, and the filter unit 312 may generate decoded video data based on the syntax elements extracted from the bitstream. The filter unit 312 may be configured to perform any of the NN-based video coding techniques described above.

[0198] Typically, the video decoder 300 reconstructs a picture on a block-by-block basis. The video decoder 300 may perform a reconstruction operation on each block individually (where the block currently being reconstructed (ie, decoded) may be referred to as the "current block").

[0199] The entropy decoding unit 302 may entropy decode syntax elements defining the quantized transform coefficients of the quantized transform coefficient block, as well as transform information such as a quantization parameter (QP) and / or a transform mode indication. The inverse quantization unit 306 may use the QP associated with the quantized transform coefficient block to determine a quantization level and, similarly, determine an inverse quantization level to be applied by the inverse quantization unit 306. The inverse quantization unit 306 may, for example, perform a bitwise left shift operation to inverse quantize the quantized transform coefficients. The inverse quantization unit 306 may thereby form a transform coefficient block comprising the transform coefficients.

[0200] After inverse quantization unit 306 forms a transform coefficient block, inverse transform processing unit 308 may apply one or more inverse transforms to the transform coefficient block to generate a residual block associated with the current block. For example, inverse transform processing unit 308 may apply an inverse DCT, an inverse integer transform, an inverse Karhunen-Loeve transform (KLT), an inverse rotational transform, an inverse directional transform, or another inverse transform to the transform coefficient block.

[0201] In addition, prediction processing unit 304 generates a prediction block based on the prediction information syntax element entropy decoded by entropy decoding unit 302. For example, if the prediction information syntax element indicates that the current block is inter-predicted, motion compensation unit 316 may generate a prediction block. In this case, the prediction information syntax element may indicate a reference picture in DPB 314 from which to retrieve the reference block, and a motion vector that identifies the location of the reference block in the reference picture relative to the location of the current block in the current picture. Motion compensation unit 316 may generally generate a prediction block in the same manner as for motion compensation unit 224 ( Figure 15 ) is performed in a manner substantially similar to that described in the foregoing.

[0202] As another example, if the prediction information syntax element indicates that the current block is intra-predicted, the intra-prediction unit 318 may generate a prediction block according to the intra-prediction mode indicated by the prediction information syntax element. Again, the intra-prediction unit 318 may generally generate a prediction block in the same manner as described with respect to the intra-prediction unit 226 ( Figure 15 The intra prediction process is performed in a manner substantially similar to that described in the preceding claims. The intra prediction unit 318 may retrieve data of neighboring samples of the current block from the DPB 314.

[0203] The reconstruction unit 310 may reconstruct the current block using the prediction block and the residual block. For example, the reconstruction unit 310 may add samples of the residual block to corresponding samples of the prediction block to reconstruct the current block.

[0204] Filter unit 312 may perform one or more filter operations on the reconstructed block. For example, filter unit 312 may perform a deblocking operation to reduce blocking artifacts along the edges of the reconstructed block. The operations of filter unit 312 may not necessarily be performed in all examples. Filter unit 312 may be configured to perform any of the NN-based video coding techniques described above. For example, filter unit 216 may be configured to receive a reconstructed block of a picture of video data and perform an NN-based filtering process on the reconstructed block to produce a filtered block. The NN-based filtering process includes performing multiple separable convolutions to approximate a multi-dimensional convolution.

[0205] The video decoder 300 may store the reconstructed block in the DPB 314. For example, in examples where the operation of the filter unit 312 is not performed, the reconstruction unit 310 may store the reconstructed block in the DPB 314. In examples where the operation of the filter unit 312 is performed, the filter unit 312 may store the filtered reconstructed block in the DPB 314. As described above, the DPB 314 may provide reference information to the prediction processing unit 304, such as samples of the current picture for intra-frame prediction and previously decoded pictures for subsequent motion compensation. In addition, the video decoder 300 may output a decoded picture (e.g., a decoded video) from the DPB 314 for display on a display device such as a video processor. Figure 1 Subsequent presentation on the display device 118).

[0206] In this way, video decoder 300 represents an example of a video decoding device that includes: a memory configured to store video data; and one or more processing units implemented in circuits and configured to perform NN-based video decoding, including NN-based filtering using any combination of the techniques described above.

[0207] Figure 17 1 is a flowchart illustrating an example method for encoding a current block according to the techniques of the present disclosure. The current block may be or may include a current CU. Although relative to the video encoder 200 ( Figure 1 and 15 ), but it should be understood that other devices may be configured to perform the same Figure 17 A similar approach to the one in the previous section.

[0208] In this example, the video encoder 200 first predicts the current block (350). For example, the video encoder 200 may form a prediction block for the current block. The video encoder 200 may then calculate a residual block for the current block (352). To calculate the residual block, the video encoder 200 may calculate the difference between the original, unencoded block and the prediction block for the current block. The video encoder 200 may then transform the residual block and quantize the transform coefficients of the residual block (354). Next, the video encoder 200 may scan the quantized transform coefficients of the residual block (356). During or after the scan, the video encoder 200 may entropy encode the transform coefficients (358). For example, the video encoder 200 may encode the transform coefficients using CAVLC or CABAC. The video encoder 200 may then output entropy encoded data for the block (360).

[0209] Figure 18 1 is a flowchart illustrating an example method for decoding a current block according to the techniques of the present disclosure. The current block may be or may include a current CU. Although relative to the video decoder 300 ( Figure 1 and 16 ), but it should be understood that other devices may be configured to perform the same Figure 18 A similar approach to the one in the previous section.

[0210] The video decoder 300 may receive entropy-encoded data for a current block, such as entropy-encoded prediction information and entropy-encoded data of transform coefficients for a residual block corresponding to the current block (370). The video decoder 300 may entropy decode the entropy-encoded data to determine the prediction information for the current block and reproduce the transform coefficients of the residual block (372). The video decoder 300 may predict the current block (374), for example, using an intra-frame or inter-frame prediction mode as indicated by the prediction information of the current block to calculate a prediction block for the current block. The video decoder 300 may then inverse scan the reproduced transform coefficients (376) to create a block of quantized transform coefficients. The video decoder 300 may then inverse quantize the transform coefficients and apply an inverse transform to the transform coefficients to produce a residual block (378). The video decoder 300 may ultimately decode the current block (380) by combining the prediction block and the residual block.

[0211] Figure 19 is a flowchart illustrating an example method for coding video data using NN-based filters, in accordance with techniques of this disclosure. Figure 19 The technique may be performed by one or more units of the video encoder 200 and the video decoder 300, including the filter unit 216 ( Figure 15 ) and filter unit 312 ( Figure 16 ).

[0212] In one example, the video encoder 200 and the video decoder 300 can be configured to receive a picture of video data (1900) and reconstruct a block of the picture of the video data to generate a reconstructed block (1910). The video encoder 200 and the video decoder 300 can also be configured to perform a neural network (NN)-based filtering process on the reconstructed block to generate a filtered block, wherein the NN-based filtering process includes performing multiple separable convolutions to approximate a multidimensional convolution (1920). In one example, the multidimensional convolution has a kernel size of n1xn2 in the spatial dimension and a size of K in the depth dimension. As a specific example, the multidimensional convolution can be a 3x3 convolution, but other sizes are also possible.

[0213] In one example of performing multiple separable convolutions to approximate a multi-dimensional convolution, the video encoder 200 and the video decoder 300 are configured to perform a first 1x1 convolution, perform a first separable convolution of multiple separable convolutions on the output of the first 1x1 convolution, perform a second separable convolution of multiple separable convolutions on the output of the first separable convolution, and perform a second 1x1 convolution on the output of the second separable convolution.

[0214] The video encoder 200 and the video decoder 300 can perform multiple separable convolutions to approximate multi-dimensional convolutions in a backbone block of the NN-based filter process. The backbone block can be one of a residual block, a filter block, or a focus residual block.

[0215] In one example, when the backbone block is a residual block, performing multiple separable convolutions to approximate a multi-dimensional convolution includes receiving an input at the residual block, performing a 1x1xKxM convolution on the input, performing a PReLU layer on the output of the 1x1xKxM convolution, performing a 1x1xMxR convolution on the output of the PReLU layer, performing a 3x1xRxR separable convolution on the output of the 1x1xMxR convolution, performing a 1x3xRxR separable convolution on the output of the 3x1xRxR separable convolution, and performing a 1x1xRxK convolution on the output of the 1x3xRxR separable convolution. In one example, the 1x1xMxR convolution is a fusion of the 1x1xMxK convolution and the 1x1xKxR convolution.

[0216] In another example, where the backbone block is a residual block, performing multiple separable convolutions to approximate a multidimensional convolution includes receiving an input at the residual block, performing a 1x1xKxM convolution on the input, performing a PReLU layer on the output of the 1x1xKxM convolution, performing a 1x1xMxK convolution on the output of the PReLU layer, performing a 3x1xKxR separable convolution on the output of the 1x1xMxK convolution, performing a 1x3xRxK separable convolution on the output of the 3x1xKxR separable convolution, and performing a 1x1xRxK convolution on the output of the 1x3xRxR separable convolution.

[0217] In some examples, the NN-based filtering process includes cascaded application of stem blocks. In other examples, the NN-based filtering process includes cascaded application of stem blocks applied in two or more parallel processing branches.

[0218] In any of the above examples, the video encoder 200 and the video decoder 300 can apply an element-wise activation process as part of the multi-dimensional convolution. In one example, the element-wise activation process is controlled parametrically.

[0219] In another example, to perform multiple separable convolutions to approximate multi-dimensional convolution, the video encoder 200 and the video decoder 300 are configured to perform a first convolution of size n1×1 and perform a second convolution of size 1×n2 on the output of the first convolution.

[0220] In any of the above examples, performing multiple separable convolutions to approximate a multidimensional convolution may include performing multiple separable convolutions to approximate a multidimensional convolution in one or more of a feature extraction portion, a fusion block, a transition block, a backbone block, or a tail portion of a NN-based filter process.

[0221] In any of the above examples, video encoder 200 and video decoder 300 may be configured to use a decoded picture including a filtered block as a reference for predicting other coded pictures.

[0222] The following numbered clauses illustrate one or more aspects of the devices and techniques described in this disclosure.

[0223] Aspect 1A. A method of encoding video data, the method comprising: receiving a picture of the video data; reconstructing the picture of the video data; and performing a neural network (NN)-based filtering process on the reconstructed picture of the video data, wherein the NN-based filtering process comprises approximating a 3x3 convolution using one or more separable convolutions.

[0224] Aspect 2A. The method of any of Aspects 1A, wherein transcoding comprises decoding, and wherein reconstructing comprises decoding.

[0225] Aspect 3A. The method according to any of Aspects 1A, wherein decoding comprises encoding.

[0226] Aspect 4A. An apparatus for decoding video data, the apparatus comprising one or more means for performing the method according to any of aspects 1A-3A.

[0227] Aspect 5A. The apparatus of aspect 4A, wherein the one or more units include one or more processors implemented in circuitry.

[0228] Aspect 6A. The apparatus of any of aspects 4A and 5A, further comprising a memory for storing video data.

[0229] Aspect 7A. The apparatus of any of clauses 4A-6A, further comprising: a display configured to display the decoded video data.

[0230] Aspect 8A. The device of any of clauses 4A-7A, wherein the device comprises one or more of a camera, a computer, a mobile device, a broadcast receiver device, or a set-top box.

[0231] Aspect 9A. The apparatus of any of aspects 4A-8A, wherein the apparatus comprises a video decoder.

[0232] Aspect 10A. The apparatus of any of aspects 4A-9A, wherein the apparatus comprises a video encoder.

[0233] Aspect 11A. A computer-readable storage medium having stored thereon instructions that, when executed, cause one or more processors to perform the method of any of Aspects 1A-3A.

[0234] Aspect 1B. A method of decoding video data, the method comprising: receiving a picture of video data; reconstructing a block of the picture of the video data to produce a reconstructed block; and performing a neural network (NN)-based filtering process on the reconstructed block to produce a filtered block, wherein the NN-based filtering process comprises performing a plurality of separable convolutions to approximate a multidimensional convolution.

[0235] Aspect 2B. The method according to Aspect 1B, wherein the multi-dimensional convolution has a kernel size of n1xn2 in the spatial dimension and a size of K in the depth dimension.

[0236] Aspect 3B. The method of any one of Aspects 1B-2B, wherein performing a plurality of separable convolutions to approximate a multidimensional convolution comprises: performing a first 1x1 convolution; performing a first separable convolution of the plurality of separable convolutions on an output of the first 1x1 convolution; performing a second separable convolution of the plurality of separable convolutions on an output of the first separable convolution; and performing a second 1x1 convolution on an output of the second separable convolution.

[0237] Aspect 4B. The method of aspect 3B, wherein performing the plurality of separable convolutions to approximate the multi-dimensional convolution comprises performing the plurality of separable convolutions to approximate the multi-dimensional convolution in a backbone block of the NN-based filter process.

[0238] Aspect 5B. The method according to aspect 4B, wherein the backbone block is one of a residual block, a filter block, or a focused residual block.

[0239] Aspect 6B. The method of aspect 4B, wherein the backbone block is a residual block, and wherein performing a plurality of separable convolutions to approximate a multi-dimensional convolution comprises: receiving an input at the residual block; performing a 1x1xKxM convolution on the input; performing a PReLU layer on the output of the 1x1xKxM convolution; performing a 1x1xMxR convolution on the output of the PReLU layer; performing a 3x1xRxR separable convolution on the output of the 1x1xMxR convolution; performing a 1x3xRxR separable convolution on the output of the 3x1xRxR separable convolution; and performing a 1x1xRxK convolution on the output of the 1x3xRxR separable convolution.

[0240] Aspect 7B. The method according to Aspect 6B, wherein the 1x1xMxR convolution is a fusion of a 1x1xMxK convolution and a 1x1xKxR convolution.

[0241] Aspect 8B. The method of aspect 4B, wherein the backbone block is a residual block, and wherein performing a plurality of separable convolutions to approximate a multi-dimensional convolution comprises: receiving an input at the residual block; performing a 1x1xKxM convolution on the input; performing a PReLU layer on the output of the 1x1xKxM convolution; performing a 1x1xMxK convolution on the output of the PReLU layer; performing a 3x1xKxR separable convolution on the output of the 1x1xMxK convolution; performing a 1x3xRxK separable convolution on the output of the 3x1xKxR separable convolution; and performing a 1x1xRxK convolution on the output of the 1x3xRxK separable convolution.

[0242] Aspect 9B. The method of any of aspects 4B-8B, wherein the NN-based filter process comprises a cascaded application of backbone blocks.

[0243] Aspect 10B. The method of any of aspects 4B-8B, wherein the NN-based filter process comprises cascaded application of the backbone blocks applied in two or more parallel processing branches.

[0244] Aspect 11B. The method according to any of aspects 4B-10B, further comprising: applying an element-wise activation process as part of a multi-dimensional convolution.

[0245] Aspect 12B. The method of aspect 11B, wherein the element-wise activation process is controlled parametrically.

[0246] Aspect 13B. The method of any of Aspects 1B-12B, wherein performing a plurality of separable convolutions to approximate a multi-dimensional convolution comprises: performing a first convolution of size n1x1; and performing a second convolution of size 1xn2 on an output of the first convolution.

[0247] Aspect 14B. The method of any one of Aspects 1B-13B, wherein performing a plurality of separable convolutions to approximate a multi-dimensional convolution comprises performing a plurality of separable convolutions to approximate a multi-dimensional convolution in one or more of a feature extraction portion, a fusion block, a transition block, a backbone block, or a tail portion of the NN-based filtering process.

[0248] Aspect 15B. The method of any of aspects 1B-14B, wherein coding comprises decoding, and wherein the method further comprises using a decoded picture comprising the filtered block as a reference for predicting other coded pictures.

[0249] Aspect 16B. The method of any of aspects 1B-14B, wherein decoding comprises encoding, and wherein the method further comprises: capturing a picture of the video data using a camera.

[0250] Aspect 17B. A device configured to code video data, the device comprising: a memory configured to store a picture of the video data; and processing circuitry in communication with the memory, the processing circuitry configured to: receive the picture of the video data; reconstruct a block of the picture of the video data to produce a reconstructed block; and perform a neural network (NN)-based filtering process on the reconstructed block to produce a filtered block, wherein the NN-based filtering process comprises performing a plurality of separable convolutions to approximate a multidimensional convolution.

[0251] Aspect 18B. The apparatus of aspect 17B, wherein the multi-dimensional convolution is a 3x3 convolution.

[0252] Aspect 19B. The apparatus of any of Aspects 17B-18B, wherein, to perform a plurality of separable convolutions to approximate a multidimensional convolution, the processing circuit is further configured to: perform a first 1x1 convolution; perform a first separable convolution of the plurality of separable convolutions on an output of the first 1x1 convolution; perform a second separable convolution of the plurality of separable convolutions on an output of the first separable convolution; and perform a second 1x1 convolution on an output of the second separable convolution.

[0253] Aspect 20B. The apparatus of aspect 19B, wherein, to perform multiple separable convolutions to approximate the multi-dimensional convolution, the processing circuit is further configured to: perform multiple separable convolutions to approximate the multi-dimensional convolution in a backbone block of the NN-based filter process.

[0254] Aspect 21B. The apparatus of aspect 20B, wherein the backbone block is one of a residual block, a filter block, or an attention residual block.

[0255] Aspect 22B. The apparatus of Aspect 20B, wherein the backbone block is a residual block, and wherein, to perform a plurality of separable convolutions to approximate a multi-dimensional convolution, the processing circuit is further configured to: receive an input at the residual block; perform a 1x1xKxM convolution on the input; perform a PReLU layer on the output of the 1x1xKxM convolution; perform a 1x1xMxR convolution on the output of the PReLU layer; perform a 3x1xRxR separable convolution on the output of the 1x1xMxR convolution; perform a 1x3xRxR separable convolution on the output of the 3x1xRxR separable convolution; and perform a 1x1xRxK convolution on the output of the 1x3xRxR separable convolution.

[0256] Aspect 23B. The apparatus of aspect 22B, wherein the 1x1xMxR convolution is a fusion of a 1x1xMxK convolution and a 1x1xKxR convolution.

[0257] Aspect 24B. The apparatus of any of Aspects 20B-23B, wherein the NN-based filter process comprises a cascaded application of backbone blocks.

[0258] Aspect 25B. The apparatus of any of Aspects 20B-23B, wherein the NN-based filter process comprises cascaded application of the backbone blocks applied in two or more parallel processing branches.

[0259] Aspect 26B. The apparatus of any of Aspects 20B-25B, wherein the processing circuit is further configured to apply an element-wise activation process as part of the multidimensional convolution.

[0260] Aspect 27B. The apparatus of aspect 26B, wherein the element-wise activation process is controlled parametrically.

[0261] Aspect 28B. The apparatus of any one of Aspects 17B-27B, wherein, to perform the plurality of separable convolutions to approximate the multidimensional convolution, the processing circuit is further configured to: perform the plurality of separable convolutions to approximate the multidimensional convolution in one or more of a feature extraction portion, a fusion block, a transition block, a backbone block, or a tail portion of the NN-based filtering process.

[0262] Aspect 29B. The apparatus of any of aspects 17B-28B, wherein the apparatus is configured to decode video data, and wherein the processing circuitry is further configured to use a decoded picture including the filtered block as a reference for predicting other coded pictures.

[0263] Aspect 30B. The apparatus of any of aspects 17B to 28B, wherein the apparatus is configured to encode video data, and wherein the apparatus further comprises a camera configured to capture pictures of the video data.

[0264] It is to be appreciated that, depending on the examples, certain actions or events of any of the techniques described herein may be performed in a different order, may be added, combined, or omitted entirely (e.g., not all described actions or events are necessary for implementation of the techniques). Furthermore, in certain examples, actions or events may be performed concurrently (e.g., through multithreading, interrupt handling, or multiple processors) rather than sequentially.

[0265] In one or more examples, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions may be stored or transmitted as one or more instructions or codes on a computer-readable medium and executed by a hardware-based processing unit. Computer-readable media may include: computer storage media, which corresponds to tangible media such as data storage media; or communication media, including any media that facilitates the transfer of a computer program from one place to another, for example, according to a communication protocol. In this manner, computer-readable media may generally correspond to: (1) a non-transitory tangible computer-readable storage medium, or (2) a communication medium such as a signal or carrier wave. Data storage media may be any available medium that can be accessed by one or more computers or one or more processors to retrieve instructions, codes, and / or data structures to implement the techniques described in this disclosure. A computer program product may include computer-readable media.

[0266] By way of example and not limitation, such computer-readable storage media may include one or more of the following: RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage device, flash memory, or any other medium that can be used to store desired program code in the form of instructions or data structures and that can be accessed by a computer. In addition, any connection is appropriately referred to as a computer-readable medium. For example, if software is sent from a website, server or other remote source using a coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL) or wireless technologies such as infrared, wireless and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL or wireless technologies such as infrared, wireless and microwave are also included in the definition of medium. However, it should be understood that computer-readable storage media and data storage media do not include connections, carrier waves, signals or other temporary media, but are directed to non-temporary tangible storage media. Disks and optical disks used herein include compact discs (CDs), laser discs, optical discs, digital versatile discs (DVDs), floppy disks and Blu-ray discs, where disks typically reproduce data magnetically, while optical discs typically reproduce data optically using lasers. The above combinations should also be included within the scope of computer-readable media.

[0267] Instructions may be executed by one or more processors, such as one or more DSPs, general-purpose microprocessors, ASICs, FPGAs, or other equivalent integrated or discrete logic circuits. Thus, as used herein, the terms "processor" and "processing circuitry" may refer to any of the aforementioned structures or any other structure suitable for implementing the techniques described herein. Additionally, in some aspects, the functionality described herein may be provided within dedicated hardware and / or software modules configured for encoding and decoding, or incorporated into a combined codec. Similarly, the techniques may be fully implemented in one or more circuits or logic elements.

[0268] The techniques of the present disclosure may be implemented in a variety of devices or apparatuses, including wireless handsets, integrated circuits (ICs), or a set of ICs (e.g., chipsets). Various components, modules, or units are described in this disclosure to emphasize functional aspects of devices configured to perform the disclosed techniques, but they do not necessarily need to be implemented by different hardware units. Rather, as described above, the various units may be combined in a codec hardware unit, or provided by a collection of interoperable hardware units, including one or more processors as described above in combination with appropriate software and / or firmware.

[0269] Various examples have been described. These and other examples are within the scope of the following claims.

Claims

1. A method for decoding video data, the method comprising: Receive a picture of the video data; reconstructing blocks of the picture of video data to generate reconstructed blocks; as well as A neural network (NN) based filtering process is performed on the reconstructed block to generate a filtered block, wherein the NN based filtering process includes performing a plurality of separable convolutions to approximate a multi-dimensional convolution.

2. The method according to claim 1, wherein The multidimensional convolution has a kernel size of n1xn2 in the spatial dimension and a size of K in the depth dimension.

3. The method according to claim 1, wherein Performing the plurality of separable convolutions to approximate the multidimensional convolution includes: Perform the first 1x1 convolution; Performing a first separable convolution on an output of the first 1x1 convolution; performing a second separable convolution of the plurality of separable convolutions on an output of the first separable convolution; and A second 1x1 convolution is performed on the output of the second separable convolution.

4. The method according to claim 3, wherein: Performing the plurality of separable convolutions to approximate the multidimensional convolution includes: The plurality of separable convolutions are performed in a backbone block of the NN-based filtering process to approximate the multi-dimensional convolution.

5. The method according to claim 4, wherein The backbone block is one of a residual block, a filter block, or a residual block of interest.

6. The method according to claim 4, wherein: The backbone block is a residual block, and wherein performing the plurality of separable convolutions to approximate the multidimensional convolution comprises: receiving an input at the residual block; Perform a 1x1xKxM convolution on the input; Perform a PReLU layer on the output of the 1x1xKxM convolution; Perform a 1x1xMxR convolution on the output of the PReLU layer; Performing a 3x1xRxR separable convolution on the output of the 1x1xMxR convolution; performing a 1x3xRxR separable convolution on the output of the 3x1xRxR separable convolution; and A 1x1xRxK convolution is performed on the output of the 1x3xRxR separable convolution.

7. The method according to claim 6, wherein: The 1x1xMxR convolution is a fusion of the 1x1xMxK convolution and the 1x1xKxR convolution.

8. The method according to claim 4, wherein The backbone block is a residual block, and wherein performing the plurality of separable convolutions to approximate the multidimensional convolution comprises: receiving an input at the residual block; Perform a 1x1xKxM convolution on the input; Perform a PReLU layer on the output of the 1x1xKxM convolution; Perform a 1x1xMxK convolution on the output of the PReLU layer; Perform a 3x1xKxR separable convolution on the output of the 1x1xMxK convolution; performing a 1x3xRxK separable convolution on the output of the 3x1xKxR separable convolution; and A 1x1xRxK convolution is performed on the output of the 1x3xRxR separable convolution.

9. The method according to claim 4, wherein: The NN-based filtering process includes a cascaded application of the backbone blocks.

10. The method according to claim 4, wherein: The NN-based filtering process includes cascaded application of the backbone blocks applied in two or more parallel processing branches.

11. The method according to claim 4, further comprising: An element-wise activation process is applied as part of the multidimensional convolution.

12. The method according to claim 11, wherein The element-wise activation process is controlled parametrically.

13. The method according to claim 1, wherein Performing the plurality of separable convolutions to approximate the multidimensional convolution includes: Perform a first convolution of size n1x1; and A second convolution of size 1xn2 is performed on the output of the first convolution.

14. The method according to claim 1, wherein Performing the plurality of separable convolutions to approximate the multidimensional convolution includes: The plurality of separable convolutions are performed in one or more of a feature extraction portion, a fusion block, a transition block, a backbone block, or a tail portion of the NN-based filtering process to approximate the multi-dimensional convolution.

15. The method according to claim 1, wherein Decoding includes decoding, and wherein the method further comprises: A decoded picture including the filtered block is used as a reference for predicting other coded pictures.

16. The method according to claim 1, wherein Decoding includes encoding, and wherein the method further comprises: The pictures of the video data are captured using a camera.

17. An apparatus configured to decode video data, the apparatus comprising: a memory configured to store pictures of the video data; as well as a processing circuit in communication with the memory, the processing circuit being configured to: receiving the picture of video data; reconstructing blocks of the picture of video data to generate reconstructed blocks; as well as A neural network (NN) based filtering process is performed on the reconstructed block to generate a filtered block, wherein the NN based filtering process includes performing a plurality of separable convolutions to approximate a multi-dimensional convolution.

18. The device according to claim 17, wherein The multidimensional convolution is a 3x3 convolution.

19. The device according to claim 17, wherein To perform the plurality of separable convolutions to approximate the multi-dimensional convolution, the processing circuit is further configured to: Perform the first 1x1 convolution; Performing a first separable convolution on an output of the first 1x1 convolution; performing a second separable convolution of the plurality of separable convolutions on an output of the first separable convolution; as well as A second 1x1 convolution is performed on the output of the second separable convolution.

20. The device according to claim 19, wherein To perform the plurality of separable convolutions to approximate the multi-dimensional convolution, the processing circuit is further configured to: The plurality of separable convolutions are performed in a backbone block of the NN-based filtering process to approximate the multi-dimensional convolution.

21. The device according to claim 20, wherein The backbone block is one of a residual block, a filter block, or a residual block of interest.

22. The device according to claim 20, wherein The backbone block is a residual block, and wherein, in order to perform the plurality of separable convolutions to approximate the multidimensional convolution, the processing circuit is further configured to: receiving an input at the residual block; Perform a 1x1xKxM convolution on the input; Perform a PReLU layer on the output of the 1x1xKxM convolution; Perform a 1x1xMxR convolution on the output of the PReLU layer; Performing a 3x1xRxR separable convolution on the output of the 1x1xMxR convolution; performing a 1x3xRxR separable convolution on the output of the 3x1xRxR separable convolution; and A 1x1xRxK convolution is performed on the output of the 1x3xRxR separable convolution.

23. The device according to claim 22, wherein The 1x1xMxR convolution is a fusion of the 1x1xMxK convolution and the 1x1xKxR convolution.

24. The apparatus according to claim 20, wherein The NN-based filtering process includes a cascaded application of the backbone blocks.

25. The apparatus according to claim 20, wherein The NN-based filtering process includes cascaded application of the backbone blocks applied in two or more parallel processing branches.

26. The apparatus according to claim 20, wherein The processing circuit is further configured to: An element-wise activation process is applied as part of the multidimensional convolution.

27. The device according to claim 26, wherein The element-wise activation process is controlled parametrically.

28. The apparatus according to claim 17, wherein To perform the plurality of separable convolutions to approximate the multi-dimensional convolution, the processing circuit is further configured to: The plurality of separable convolutions are performed in one or more of a feature extraction portion, a fusion block, a transition block, a backbone block, or a tail portion of the NN-based filtering process to approximate the multi-dimensional convolution.

29. The apparatus according to claim 17, wherein The apparatus is configured to decode video data, and wherein the processing circuit is further configured to: A decoded picture including the filtered block is used as a reference for predicting other coded pictures.

30. The apparatus of claim 17, wherein: The apparatus is configured to encode video data, and wherein the apparatus further comprises: A camera is configured to capture the picture of the video data.

Images