Neural network architecture for extremely low-complexity in-loop filters in video decoding

By using a NN-based filter with a restricted header block in video decoding, the output of the input channel is extracted and fused independently, solving the problem of high complexity in CNN filtering technology and achieving a performance improvement in low-complexity video decoding.

CN123125884APending Publication Date: 2026-11-17QUALCOMM INC
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202580021364.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2025-02-26
Filing Date
2025-02-27
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing loop filtering techniques based on convolutional neural networks (CNNs) are highly complex in video decoding, resulting in significant processing power requirements and making them difficult to implement effectively on resource-constrained devices.

Method used

By employing a neural network (NN)-based filter with a restricted head block, the output channels of multiple input channels are extracted independently and fused into a single output, reducing the processing burden on the processor and achieving a decoding performance improvement with extremely low complexity.

Benefits of technology

It reduces the processing complexity of video decoding equipment, improves video decoding performance, enables more devices to perform NN-based filtering, and improves the overall efficiency of video decoding.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123125884A_ABST
    Figure CN123125884A_ABST
Patent Text Reader

Abstract

An example device includes one or more memories and one or more processors. The one or more processors are configured to: receive video data indicating an image; reconstruct blocks of the video data to generate reconstructed blocks; and perform a neural network (NN)-based filtering process on the reconstructed blocks. The NN-based filtering process includes applying an NN-based filter that includes a restricted header block. The restricted header block includes multiple input channels. The restricted header block is configured to independently extract a corresponding number of output channels for each of the multiple input channels. The restricted header block is configured to fuse the outputs associated with each of the multiple input channels into a single output. The sum of the corresponding number of output channels is less than or equal to the number of fused output channels of the single output.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] This application claims priority to U.S. Patent Application No. 19 / 063,511, filed February 26, 2025, and U.S. Provisional Patent Application No. 63 / 567,585, filed March 20, 2024, the entire contents of each of which are incorporated herein by reference. U.S. Patent Application No. 19 / 063,511, filed February 26, 2025, claims the benefit of U.S. Provisional Patent Application No. 63 / 567,585, filed March 20, 2024. Technical Field

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

[0003] Digital video capabilities can be incorporated into a wide variety of devices, including digital televisions, digital live broadcast systems, wireless broadcasting systems, personal digital assistants (PDAs), laptops or desktop computers, tablets, e-book readers, digital cameras, digital recording devices, digital media players, video game devices, video game consoles, cellular or satellite wireless phones (so-called "smartphones"), video conferencing equipment, video streaming devices, and more. Digital video devices implement video decoding technologies, such as those defined by MPEG-2, MPEG-4, ITU-T H.263, ITU-T H.264 / MPEG-4 (Part 10, Advanced Video Decoding (AVC)), ITU-T H.265 / High Efficiency Video Decoding (HEVC), ITU-T H.266 / Variety Video Decoding (VVC) and extensions to these standards, as well as proprietary video codecs / formats such as AOMedia Video1 (AV1) developed by the Open Media Alliance. By implementing such video decoding technologies, video devices can more efficiently send, receive, encode, decode, and / or store digital video information.

[0004] Video decoding techniques include spatial (intra-picture) prediction and / or temporal (inter-picture) prediction to reduce or remove redundancy inherent in video sequences. For block-based video decoding, video slices (e.g., video pictures or portions of video pictures) can be divided into video blocks, which may also be referred to as decoding tree units (CTUs), decoding units (CUs), and / or decoding nodes. Video blocks in a slice after intra-frame decoding (I) of a picture are encoded using spatial prediction relative to reference samples in adjacent blocks within the same picture. Video blocks in a slice after inter-frame decoding (P or B) of a picture can use spatial prediction relative to reference samples in adjacent blocks within the same picture or temporal prediction relative to reference samples in other reference pictures. A picture may be referred to as a frame, and a reference picture may be referred to as a reference frame. Summary of the Invention

[0005] This disclosure generally describes techniques for video decoding. Specifically, it describes methods, techniques, and apparatuses for improving decoding performance under constraints of complexity and / or memory requirements. An example convolutional neural network (CNN)-based in-loop filtering architecture is described. Decoding performance improvements with extremely low-complexity architectures are achieved through improved input context feature extraction and data fusion, transformation, and higher-rate luminance / chrominance network segmentation. While the example techniques described herein relate to CNN-assisted in-loop filtering, such techniques are applicable to any CNN-based video decoding tool that consumes input data with certain statistical properties. The techniques disclosed herein can be used in the context of advanced video codecs, such as extensions to VVC, next-generation video decoding standards, and / or any other video codec.

[0006] In one example, a method for decoding video data includes: receiving video data indicating an image; reconstructing blocks of the video data to generate reconstructed blocks; and performing a neural network (NN)-based filtering process on the reconstructed blocks to generate filtered blocks, wherein the NN-based filtering process includes applying an NN-based filter that includes a restricted header block having multiple input channels and configured to independently extract a corresponding number of output channels for each of the multiple input channels and fuse the outputs associated with each of the multiple input channels into a single output, wherein the sum of the corresponding number of output channels is less than or equal to the number of fused output channels of the single output.

[0007] In another example, an apparatus configured to decode video data includes: one or more memories configured to store images of the video data; and one or more processors communicating with the one or more memories, the one or more processors being configured to: receive video data indicating the images; reconstruct blocks of the video data to generate reconstructed blocks; and perform a neural network (NN)-based filtering process on the reconstructed blocks to generate filtered blocks, wherein the NN-based filtering process includes applying an NN-based filter that includes a restricted head block having multiple input channels and being configured to independently extract a corresponding number of output channels for each of the multiple input channels and fuse the outputs associated with each of the multiple input channels into a single output, wherein the sum of the corresponding number of output channels is less than or equal to the number of fused output channels of the single output.

[0008] In another example, this disclosure describes an apparatus for encoding video data, the apparatus comprising: one or more memories configured to store images of video data; and one or more processors communicating with the one or more memories, the one or more processors configured to: determine video data indicating the images; reconstruct blocks of the video data to generate reconstructed blocks; and perform a neural network (NN)-based filtering process on the reconstructed blocks to generate filtered blocks, wherein the NN-based filtering process includes applying an NN-based filter comprising a restricted head block having a plurality of input channels and configured to independently extract a corresponding number of output channels for each of the plurality of input channels and fuse the outputs associated with each of the plurality of input channels into a single output, wherein the sum of the corresponding number of output channels is less than or equal to the number of fused output channels of the single output.

[0009] Details of one or more examples are set forth in the accompanying drawings and the following description. Other features, objects, and advantages will be apparent from the description, drawings, and claims. Attached Figure Description

[0010] Figure 1 This is a block diagram illustrating an example video encoding and decoding system that can perform the techniques of this disclosure.

[0011] Figure 2 This is a block diagram illustrating an example of a hybrid video decoding framework.

[0012] Figure 3 An example hierarchical prediction structure with a group of pictures (GOP) of size 16 is shown.

[0013] Figure 4 This is a block diagram illustrating an example of a Convolutional Neural Network (CNN) based filter with four layers.

[0014] Figure 5 This is a block diagram illustrating an example CNN-based filter with padded input samples and supplementary data.

[0015] Figure 6 This is a block diagram illustrating another example of a CNN-based filter with padded input samples and supplementary data.

[0016] Figure 7 This is an example Figure 6 A conceptual diagram of attention residual blocks.

[0017] Figure 8 This is a block diagram illustrating another example of a CNN-based filter with padded input samples and supplementary data.

[0018] Figure 9 This is a block diagram illustrating a simplified CNN-based filter architecture with padded input samples and supplementary data.

[0019] Figure 10 This is an example Figure 9 A conceptual diagram of an example residual block structure.

[0020] Figure 11 This is a block diagram illustrating another example of a simplified CNN-based filter architecture with padded input samples and supplementary data.

[0021] Figure 12 This is an example Figure 11 A conceptual diagram of an example filter block structure.

[0022] Figure 13 This is a block diagram illustrating an example CNN decomposition architecture.

[0023] Figure 14 This is a block diagram illustrating an example multi-scale feature extraction backbone network with two-component convolution.

[0024] Figure 15 This is a block diagram illustrating an example unified filter with a joint model (joint luminance and chrominance).

[0025] Figure 16 This is a block diagram illustrating the example main block.

[0026] Figure 17 This is a block diagram illustrating an example unified filter with separate luminance and chrominance models (luminance).

[0027] Figure 18 This is a block diagram illustrating an example unified filter with separate luminance and chrominance models (chrominance).

[0028] Figure 19 This is a block diagram illustrating an example low-complexity in-loop filter network architecture according to one or more aspects of this disclosure.

[0029] Figure 20 This is a block diagram illustrating example backbone blocks according to one or more aspects of this disclosure.

[0030] Figure 21 This is a block diagram illustrating an example of regular clustering (CP) decomposition.

[0031] Figure 22 This is a block diagram illustrating another example of an in-loop filter network architecture with luminance / chrominance separation.

[0032] Figure 23 It is an example backbone block based on one or more aspects of this disclosure.

[0033] Figure 24 This is a block diagram illustrating another example in-loop filter network architecture with luminance / chrominance segmentation according to one or more aspects of this disclosure.

[0034] Figure 25 It is an example backbone block based on one or more aspects of this disclosure.

[0035] Figure 26 This is a block diagram illustrating an example low-complexity backbone block according to one or more aspects of this disclosure.

[0036] Figure 27 This is a block diagram illustrating another example of a low-complexity backbone block according to one or more aspects of this disclosure.

[0037] Figure 28 This is a block diagram illustrating a specific implementation of an example fusion block according to one or more aspects of this disclosure.

[0038] Figure 29 This is a block diagram illustrating an example restricted header block design according to one or more aspects of this disclosure.

[0039] Figure 30 This is a block diagram illustrating an example sequential application and low-complexity transformation block that incorporates one or more aspects of this disclosure.

[0040] Figure 31 This is a block diagram illustrating an example neural network (NN)-based in-loop filter architecture with luminance / chrominance segmentation according to one or more aspects of this disclosure.

[0041] Figure 32 This is a flowchart illustrating an example of an NN-based in-loop filter technique according to one or more aspects of this disclosure.

[0042] Figure 33 This is a block diagram illustrating an example video encoder that can perform the techniques of this disclosure.

[0043] Figure 34 This is a block diagram illustrating an example video decoder that can perform the techniques of this disclosure.

[0044] Figure 35 This is a flowchart illustrating an example method for encoding the current block according to the technology of this disclosure.

[0045] Figure 36 This is a flowchart illustrating an example method for decoding the current block according to the technology of this disclosure. Detailed Implementation

[0046] Video coding is typically a lossy process. For example, during video coding, blocks of video data are encoded using quantization and transform. Generally, quantization of values ​​involves reducing several least significant bits of those values, which is usually irreversible. This bit reduction is generally performed in a way that avoids detectable loss. However, sometimes such loss can lead to detectable artifacts in the video data, such as block artifacts.

[0047] Filtering can be applied to decoded and / or reconstructed video data to enhance it, thus improving the output video data. For example, filtering can compensate for block artifacts or other losses in video data. Research has shown that neural network (NN)-based filtering techniques are highly capable of improving decoded and / or reconstructed video data. NN-based filtering techniques can be highly complex and require significant processing power to execute effectively.

[0048] This disclosure describes decoding performance improvement techniques applicable to neural network-based filtering. Applying these techniques can reduce the processing performed by one or more processors to execute neural network-based filtering. In this way, the techniques of this disclosure can improve the performance of video decoding devices. Similarly, these techniques enable more devices to perform neural network-based filtering, thereby substantially improving the field of video decoding.

[0049] For example, according to the technology of this disclosure, a video decoder may be configured to: receive images of video data; reconstruct blocks of the images of the video data to generate reconstructed blocks; and perform a neural network (NN)-based filtering process on the reconstructed blocks to generate filtered blocks, wherein the NN-based filtering process includes applying an NN-based filter that includes a restricted header block having multiple input channels and being configured to independently extract a corresponding number of output channels for each of the multiple input channels and fuse the outputs associated with each of the multiple input channels into a single output, wherein the sum of the corresponding number of output channels is less than or equal to the number of fused output channels of the single output.

[0050] Figure 1 This is a block diagram illustrating an example video encoding and decoding system 100 capable of performing the techniques of this disclosure. The techniques of this disclosure generally relate to decoding (encoding and / or decoding) video data. Generally speaking, video data includes any data used for processing video. Thus, video data may include unencoded raw video, encoded video, decoded (e.g., reconstructed) video, and video metadata, such as signaling data.

[0051] like Figure 1As shown, 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. Specifically, source device 102 provides video data to destination device 116 via computer-readable medium 110. Source device 102 and destination device 116 can be or may include any of a wide range of devices, such as desktop computers, laptop computers, mobile devices, tablet computers, set-top boxes, mobile phones (such as smartphones), televisions, cameras, display devices, digital media players, video game consoles, video streaming devices, broadcast receiver devices, etc. In some cases, source device 102 and destination device 116 may be configured for wireless communication and are therefore referred to as wireless communication devices.

[0052] exist Figure 1 In the example, source device 102 includes a video source 104, memory 106, video encoder 200, and output interface 108. Destination device 116 includes an input interface 122, video decoder 300, memory 120, and display device 118. According to this 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 in-loop filtering based on neural networks (NNs). 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 may include other components or arrangements. For example, source device 102 may receive video data from an external video source such as an external camera. Similarly, destination device 116 may interface with an external display device instead of including an integrated display device.

[0053] like Figure 1 The system 100 shown is merely an example. Generally, any digital video encoding and / or decoding device can perform techniques for in-loop filtering based on neural networks (NNs). Source device 102 and destination device 116 are merely examples of such decoding devices, where source device 102 generates decoded video data for transmission to destination device 116. This disclosure refers to a “decoding” device as a device that performs the decoding (e.g., encoding and / or decoding) of data. Thus, video encoder 200 and video decoder 300 represent examples of decoding devices, specifically, a video encoder and a video decoder, respectively. In some examples, source device 102 and destination device 116 may operate in a substantially symmetrical manner, such that each of source device 102 and destination device 116 includes video encoding and decoding components. Therefore, system 100 may 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.

[0054] Generally, video source 104 represents the source of video data (i.e., unencoded raw video data) and provides a sequential series of pictures (also referred to as "frames") of video data to video encoder 200, which encodes the data for the pictures. Video source 104 of source device 102 may include video capture devices such as cameras, video archives containing previously captured raw video, and / or video feed interfaces for receiving video from video content providers. Alternatively, video source 104 may generate computer graphics-based data as source video, or a combination of live video, archived video, and computer-generated video. In each case, video encoder 200 encodes the captured, pre-captured, or computer-generated video data. Video encoder 200 may rearrange the pictures from the received order (sometimes referred to as "display order") to a decoding order for decoding. Video encoder 200 may generate a bitstream comprising the encoded video data. Then, the source device 102 can output the encoded video data to the computer-readable medium 110 via the output interface 108 for reception and / or retrieval by, for example, the input interface 122 of the destination device 116.

[0055] The memory 106 of source device 102 and the memory 120 of destination device 116 represent general-purpose memory. In some examples, memories 106 and 120 may store raw video data, such as raw video from video source 104 and raw decoded video data from video decoder 300. Additionally or alternatively, memories 106 and 120 may store software instructions executable by, for example, video encoder 200 and video decoder 300. Although memories 106 and 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 may also include internal memory for functionally similar or equivalent purposes. Furthermore, memories 106 and 120 may store encoded video data, such as output from video encoder 200 and input to video decoder 300. In some examples, portions of memories 106 and 120 may be allocated as one or more video buffers, for example, to store raw decoded and / or encoded video data.

[0056] Computer-readable medium 110 may represent any type of medium or device capable of transmitting encoded video data from source device 102 to destination device 116. In one example, computer-readable medium 110 represents a communication medium that enables source device 102 to directly transmit encoded video data to 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, output interface 108 may modulate the transmitted signal including the encoded video data, and input interface 122 may demodulate the received transmitted signal. The communication medium may include any wireless or wired communication medium, such as radio frequency (RF) spectrum or one or more physical transmission lines. The communication medium may 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 may include a router, switch, base station, or any other equipment that may be useful for facilitating communication from source device 102 to destination device 116.

[0057] In some examples, source device 102 can output encoded data from output interface 108 to storage device 112. Similarly, destination device 116 can access encoded data from storage device 112 via input interface 122. Storage device 112 may include any data storage medium of various distributed or locally accessed data storage media, such as hard disk drives, Blu-ray discs, DVDs, CD-ROMs, flash memory, volatile or non-volatile memory, or any other suitable digital storage medium for storing encoded video data.

[0058] In some examples, source device 102 can output encoded video data to file server 114 or another intermediate storage device that can store the encoded video data generated by source device 102. Destination device 116 can access the stored video data from file server 114 via streaming or download.

[0059] File server 114 can be any type of server device capable of storing encoded video data and sending the encoded video data to destination device 116. File server 114 may represent a web server (e.g., for a website), a server configured to provide file transfer protocol services (such as File Transfer Protocol (FTP) or FLUTE-based file delivery protocol), 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. File server 114 may additionally or alternatively implement one or more HTTP streaming protocols, such as HTTP-based Dynamic Adaptive Streaming (DASH), HTTP Live Streaming (HLS), Real-Time Streaming Protocol (RTSP), HTTP Dynamic Streaming, etc.

[0060] Destination device 116 can access encoded video data from file server 114 via any standard data connection, including an internet connection. This may include a wireless channel (e.g., Wi-Fi connection), a wired connection (e.g., digital subscriber line (DSL), cable modem, etc.), or a combination of both, suitable for accessing encoded video data stored on file server 114. Input interface 122 can be configured to operate according to any or more of the various protocols discussed above for retrieving or receiving media data from file server 114, or other such protocols for retrieving media data.

[0061] Output interface 108 and input interface 122 can represent a wireless transmitter / receiver, a modem, a wired networking component (e.g., an Ethernet card), a wireless communication component operating according to any of the various IEEE 802.11 standards, or other physical components. In examples where output interface 108 and input interface 122 include wireless components, output interface 108 and input interface 122 can be configured to transmit data, such as encoded video data, according to cellular communication standards such as 4G, 4G-LTE (Long Term Evolution), Advanced LTE, 5G, etc. In some examples where output interface 108 includes a wireless transmitter, output interface 108 and input interface 122 can be configured to comply with other wireless standards such as the IEEE 802.11 specification, the IEEE 802.15 specification (e.g., ZigBee), etc. ™ ),Bluetooth ™Standards are used to transmit data, such as encoded video data. In some examples, source device 102 and / or destination device 116 may include corresponding system-on-chip (SoC) devices. For example, source device 102 may include an SoC device for performing functions belonging to video encoder 200 and / or output interface 108, and destination device 116 may include an SoC device for performing functions belonging to video decoder 300 and / or input interface 122.

[0062] The technology disclosed herein can be applied to video decoding to support any multimedia application in a variety of multimedia applications, such as over-the-air television broadcasting, cable television transmission, satellite television transmission, internet streaming video transmission (such as HTTP-based Dynamic Adaptive Streaming (DASH)), digital video encoded onto data storage media, decoding of digital video stored on data storage media, or other applications.

[0063] The input interface 122 of the destination device 116 receives an encoded video bitstream from a computer-readable medium 110 (e.g., a communication medium, storage device 112, 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 having values ​​describing the characteristics and / or processing of video blocks or other decoded units (e.g., slices, pictures, picture groups, sequences, etc.). The display device 118 displays a decoded picture of the decoded video data to the 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.

[0064] Despite Figure 1Not shown, but in some examples, both the video encoder 200 and the video decoder 300 may be integrated with the audio encoder and / or audio decoder (e.g., audio codec), and may include appropriate MUX-DEMUX units or other hardware and / or software to process multiplexed streams that include both audio and video in a common data stream. Example audio codecs may include AAC, AC-3, AC-4, ALAC, ALS, AMBE, AMR, AMR-WB (G.722.2), AMR-WB+, aptX (various versions), ATRAC, BroadVoice (BV16, BV32), CELT, Enhanced AC-3 (E-AC-3), EVS, FLAC, G.711, G.722, G.722.1, G.722.2 (AMR-WB), G.723.1, G.726, G.728, G.729, G.729.1, GSM-FR, HE-AAC, iLBC, iSAC, LA Lyra, Monkey's Audio, MP1, MP2 (MPEG-1, 2 Audio Layer II), MP3, Musepack, Nellymoser Asao, OptimFROG, Opus, Sac, Satin, SBC, SILK, Siren 7, Speex, SVOPC, True Audio (TTA), TwinVQ, USAC, Vorbis (Ogg), WavPack and Windows Media Aud.

[0065] Both the video encoder 200 and the video decoder 300 can 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 may store instructions for the software in a suitable non-transitory computer-readable medium and execute the instructions in hardware using one or more processors to perform the technology of this disclosure. Each of the video encoder 200 and the video decoder 300 may be included in one or more encoders or decoders, either of which may be integrated as part of a combined encoder / decoder (CODEC) in the respective device. Devices including the video encoder 200 and / or the video decoder 300 may implement the video encoder 200 and / or the video decoder 300 in processing circuitry such as integrated circuits and / or microprocessors. Such devices may be wireless communication devices (such as cellular phones) or any other type of device described herein.

[0066] The video encoder 200 and video decoder 300 may operate according to a video decoding standard such as ITU-T H.265 (also known as High Efficiency Video Decoding (HEVC)) or its extensions (such as Multi-View and / or Scalable Video Decoding Extensions)). Alternatively, the video encoder 200 and video decoder 300 may operate according to other proprietary or industry standards such as ITU-T H.266 (also known as Multi-Functional Video Decoding (VVC)). In other examples, the video encoder 200 and video decoder 300 may operate according to proprietary video codecs / formats such as AOMedia Video 1 (AV1), extensions to AV1, and / or subsequent versions of AV1 (e.g., AV2)). In other examples, the video encoder 200 and video decoder 300 may operate according to other proprietary formats or industry standards. However, the technology disclosed herein is not limited to any particular decoding standard or format. Generally speaking, the video encoder 200 and the video decoder 300 can be configured to perform the techniques of this disclosure in combination with any video decoding technique that uses neural networks (such as NN-based video decoding).

[0067] Generally, the video encoder 200 and video decoder 300 perform block-based decoding of images. The term "block" typically refers to a structure that includes data to be processed (e.g., encoded, decoded, or otherwise used during encoding and / or decoding). For example, a block may include a two-dimensional matrix of samples of luminance and / or chrominance data. Generally, the video encoder 200 and video decoder 300 decode video data represented in YUV (e.g., Y, Cb, Cr) format. That is, instead of decoding the red, green, and blue (RGB) data used for images, the video encoder 200 and video decoder 300 decode both the luminance and chrominance components, where the chrominance components may include both red and blue hue chrominance components. In some examples, the video encoder 200 converts the received RGB format data to a YUV representation before encoding, and the video decoder 300 converts that YUV representation to RGB format. Alternatively, preprocessing and postprocessing units (not shown) may perform these conversions.

[0068] This disclosure generally relates to the decoding (e.g., encoding and decoding) of images to include processes of encoding or decoding data of the image. Similarly, this disclosure may relate to the decoding of blocks of images to include processes of encoding or decoding data used for the blocks (e.g., prediction and / or residual decoding). Encoded video bitstreams typically include a series of values ​​for syntax elements representing decoding decisions (e.g., decoding modes) and the partitioning of images into blocks. Therefore, references to the decoding of images or blocks should generally be understood as the decoded values ​​of the syntax elements that form the images or blocks.

[0069] HEVC defines various blocks, including decoding units (CUs), prediction units (PUs), and transform units (TUs). According to HEVC, a video decoder (such as a video encoder 200) divides the decoding tree unit (CTU) into CUs based on 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. Nodes without child nodes are called "leaf nodes," and the CU of such leaf nodes may include one or more PUs and / or one or more TUs. The video decoder may further divide the PUs and TUs. For example, in HEVC, a residual quadtree (RQT) represents the partitioning of the TU. In HEVC, the PU represents inter-frame prediction data, while the TU represents residual data. The CU after intra-frame prediction includes intra-frame prediction information, such as intra-frame mode indication.

[0070] As another example, video encoder 200 and video decoder 300 can be configured to operate according to VVC. According to VVC, the video decoder (such as video encoder 200) partitions the image into multiple CTUs. Video encoder 200 can partition the CTUs 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 CUs, PUs, and TUs in HEVC. The QTBT structure includes two levels: a first level partitioned according to quadtree partitioning, and a second level partitioned according to binary tree partitioning. The root node of the QTBT structure corresponds to a CTU. The leaf nodes of the binary tree correspond to CUs.

[0071] In the MTT partitioning structure, blocks can be divided using quadtree (QT) partitioning, binary tree (BT) partitioning, and one or more types of ternary tree (TT) partitioning (also known as triplet tree (TT)). A ternary tree or triplet tree partition is a partition in which a block is divided into three sub-blocks. In some examples, a ternary tree or triplet tree partition divides a block into three sub-blocks without dividing the original block through the center. Partition types in MTT (e.g., QT, BT, and TT) can be symmetric or asymmetric.

[0072] When operating according to the AV1 codec, the video encoder 200 and video decoder 300 can be configured to decode video data in blocks. In AV1, the largest decoded block that can be processed is called a superblock. In AV1, a superblock can be 128x128 luminance samples or 64x64 luminance samples. However, in subsequent video decoding formats (e.g., AV2), superblocks can be defined by different (e.g., larger) luminance sample sizes. In some examples, the superblock is the top level of a block quadtree. The video encoder 200 can further divide the superblock into smaller decoded blocks. The video encoder 200 can use square or non-square partitions to divide the superblock and other decoded blocks into smaller blocks. Non-square blocks can include N / 2xN blocks, NxN / 2 blocks, N / 4xN blocks, and NxN / 4 blocks. The video encoder 200 and video decoder 300 can perform separate prediction and transform processing on each decoded block within the decoded block.

[0073] AV1 also defines video data tiles. A tile is a rectangular array of superblocks that can be decoded independently of other tiles. That is, the video encoder 200 and video decoder 300 can encode and decode the decoding blocks within a tile separately without using video data from other tiles. However, the video encoder 200 and video decoder 300 can perform filtering across tile boundaries. The tile size can be uniform or non-uniform. Tile-based decoding enables parallel processing and / or multithreading in the encoder and decoder implementations.

[0074] 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 and 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).

[0075] The video encoder 200 and the video decoder 300 can be configured to use quadtree partitioning, QTBT partitioning, MTT partitioning, superblock partitioning or other partitioning structures.

[0076] In some examples, a CTU includes a decoded tree block (CTB) of luminance samples, two corresponding CTBs of chrominance samples of an image with three sample arrays, or a CTB of samples of an image decoded using three separate color planes and a syntax structure for decoding the samples. A CTB can be an NxN sample block of some value N, such that a partitioning method divides the components into CTBs. A component is an array or a single sample from one of the three arrays (luminance and two chrominance) constituting a 4:2:0, 4:2:2, or 4:4:4 color format image, or an array or a single sample constituting an array or array constituting a monochrome format image. In some examples, a decoded block is an MxN sample block of values ​​M and N, such that a partitioning method divides the CTB into decoded blocks.

[0077] Blocks (e.g., CTUs or CUs) can be grouped in various ways within an image. As an example, a brick can refer to a rectangular area of ​​a row of CTUs within a specific tile in an image. A tile can be a rectangular area of ​​CTUs within a specific tile column and a specific tile row in an image. A tile column refers to a rectangular area of ​​a CTU having a height equal to the height of the image and a width specified by syntax elements (e.g., such as in an image parameter set). A tile row refers to a rectangular area of ​​a CTU having a height specified by syntax elements (e.g., such as in an image parameter set) and a width equal to the width of the image.

[0078] In some examples, a tile can be divided into multiple bricks, each of which may include one or more CTU rows within the tile. A tile that is not divided into multiple bricks can also be called a brick. However, bricks that are a true subset of a tile cannot be called a tile. Bricks in an image can also be arranged in slices. A slice can be an integer number of bricks in an image that can be uniquely contained within a single Network Abstraction Layer (NAL) unit. In some examples, a slice includes multiple complete tiles or a continuous sequence of complete bricks that include only one tile.

[0079] This disclosure uses "NxN" and "N by N" interchangeably to refer to the sample size of a block (such as a CU or other video block) in the vertical and horizontal dimensions, for example, 16x16 samples or 16 by 16 samples. Generally, a 16x16 CU will have 16 samples in the vertical direction (y=16) and 16 samples in the horizontal direction (x=16). Similarly, an NxNCU typically has N samples in the vertical direction and N samples in the horizontal direction, where N represents a non-negative integer value. Samples in a CU can be arranged in rows and columns. Furthermore, a CU does not necessarily need to have the same number of samples in the horizontal direction as it does in the vertical direction. For example, a CU may include NxM samples, where M is not necessarily equal to N.

[0080] The video encoder 200 encodes video data representing prediction and / or residual information, as well as other information, for use in the control unit (CU). The prediction information indicates how the CU should be predicted to form a prediction block for the CU. The residual information typically represents the sample-by-sample difference between a sample of the CU before encoding and the prediction block.

[0081] To predict the Cubic Frame (CU), the video encoder 200 typically forms prediction blocks for the CU through inter-frame prediction or intra-frame prediction. Inter-frame prediction typically refers to predicting the CU from data in a previously decoded image, while intra-frame prediction typically refers to predicting the CU from data in a previously decoded image within the same frame. To perform inter-frame prediction, the video encoder 200 can use one or more motion vectors to generate prediction blocks. The video encoder 200 can typically perform a motion search to identify reference blocks that closely match the CU, for example, based on the differences between the CU and a reference block. The video encoder 200 can use sum of absolute differences (SAD), sum of squared differences (SSD), mean absolute difference (MAD), mean squared difference (MSD), or other such difference calculations to compute difference metrics to determine whether a reference block closely matches the current CU. In some examples, the video encoder 200 can use unidirectional or bidirectional prediction to predict the current CU.

[0082] Some examples of VVC also provide an affine motion compensation mode, which can be viewed as an inter-frame prediction mode. In 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.

[0083] To perform intra-frame prediction, the video encoder 200 can select an intra-frame prediction mode to generate prediction blocks. Some examples of VVC provide sixty-seven intra-frame prediction modes, including various directional modes, as well as planar and DC modes. Generally, the video encoder 200 selects an intra-frame prediction mode that describes the neighboring samples of the current block (e.g., the block of the CU) and predicts samples of the current block from them. Assuming the video encoder 200 decodes the CTU and CU in raster scan order (from left to right, from top to bottom), such samples can typically be located above, to the upper left, or to the left of the current block in the same image as the current block.

[0084] The video encoder 200 encodes data representing the prediction mode of the current block. For example, for inter-frame prediction modes, the video encoder 200 may encode data indicating which of the various available inter-frame prediction modes is used, as well as the motion information for the corresponding mode. For example, for unidirectional or bidirectional inter-frame prediction, the video encoder 200 may encode motion vectors using Advanced Motion Vector Prediction (AMVP) or merging modes. The video encoder 200 may use similar modes to encode motion vectors used for affine motion compensation modes.

[0085] AV1 includes two common techniques for encoding and decoding blocks of video data. These two common techniques are intra-frame prediction (e.g., intra-frame prediction or spatial prediction) and inter-frame prediction (e.g., inter-frame prediction or temporal prediction). In the context of AV1, when using intra-frame prediction modes to predict blocks of video data for the current frame, the video encoder 200 and video decoder 300 do not use video data from other frames of the video data. For most intra-frame prediction modes, the video encoder 200 encodes blocks of the current frame based on the difference between sample values ​​in the current block and predicted values ​​generated from reference samples in the same frame. The video encoder 200 determines the predicted values ​​generated from the reference samples based on the intra-frame prediction mode.

[0086] After prediction (such as intra-frame or inter-frame prediction for a block), the video encoder 200 can compute residual data for the block. The residual data (such as a residual block) represents the sample-wise difference between the block and the prediction block used to form the block, which is formed using the corresponding prediction mode. The video encoder 200 can apply one or more transforms to the residual block to produce transformed data in the transform domain rather than the sample domain. For example, the video encoder 200 can apply a Discrete Cosine Transform (DCT), an integer transform, a wavelet transform, or a conceptually similar transform to the residual video data. Additionally, the video encoder 200 can apply a secondary transform after the first transform, such as a Mode Correlated Inseparable Secondary Transform (MDNSST), a Signal Correlation Transform, a Karhunen-Loeve Transform (KLT), etc. The video encoder 200 produces transform coefficients after applying one or more transforms.

[0087] As noted above, after any transformation that produces the transform coefficients, the video encoder 200 may perform quantization on the transform coefficients. Quantization generally refers to a process in which the transform coefficients are quantized to reduce the amount of data used to represent them, thereby providing further compression. By performing the quantization process, the video encoder 200 may reduce the bit depth associated with some or all of the transform coefficients. For example, the video encoder 200 may round an n-bit value down 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 bit-by-bit right shift of the value to be quantized.

[0088] After quantization, the video encoder 200 can scan the transform coefficients to generate a one-dimensional vector from a two-dimensional matrix containing the quantized transform coefficients. The scan can be designed to place higher-energy (and therefore lower-frequency) transform coefficients before the vector and lower-energy (and therefore higher-frequency) transform coefficients after the vector. In some examples, the video encoder 200 can utilize a predefined scan order to scan the quantized transform coefficients to produce a serialized vector, and then entropy-encode the quantized transform coefficients of that vector. In other examples, the video encoder 200 can perform an adaptive scan. After scanning the quantized transform coefficients to form a one-dimensional vector, the video encoder 200 can entropy-encode the one-dimensional vector, for example, according to context-adaptive binary arithmetic decoding (CABAC). The video encoder 200 can also entropy-encode the values ​​of syntax elements describing metadata associated with the encoded video data, which is used by the video decoder 300 when decoding the video data.

[0089] To perform CABAC, the video encoder 200 can assign context within a context model to the symbols to be transmitted. Context may involve, for example, whether the neighboring values ​​of a symbol are zero. Probability determination can be based on the context assigned to the symbols.

[0090] The video encoder 200 may further generate syntax data for the video decoder 300, such as block-based syntax data, image-based syntax data, and sequence-based syntax data, for example, in image headers, block headers, and slice headers, or generate other syntax data such as sequence parameter sets (SPS), image parameter sets (PPS), or video parameter sets (VPS). The video decoder 300 may also decode such syntax data to determine how to decode the corresponding video data.

[0091] In this way, the video encoder 200 can generate a bitstream that includes encoded video data, such as syntax elements describing the partitioning of images into blocks (e.g., CUs) and prediction and / or residual information for the blocks. Finally, the video decoder 300 can receive the bitstream and decode the encoded video data.

[0092] Generally, the video decoder 300 performs the reverse process of 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 used for the bitstream in a manner substantially similar to but reversed by the CABAC encoding process of the video encoder 200. Syntax elements can define partitioning information for dividing a picture into CTUs and defining the CUs of each CTU according to a corresponding partitioning structure such as a QTBT structure. Syntax elements can further define prediction and residual information for video data blocks (e.g., CUs).

[0093] The residual information can be represented, for example, by quantized transform coefficients. The video decoder 300 can inversely quantize and inverse transform the quantized transform coefficients of the block to reconstruct the residual block for the block. The video decoder 300 uses a signaling 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 can then combine the prediction block and the residual block (on a sample-by-sample basis) to reconstruct the original block. The video decoder 300 can perform additional processing, such as performing a deblocking process to reduce visual artifacts along the block boundaries.

[0094] This disclosure may generally relate to "signaling" certain information (such as syntax elements). The term "signaling" can generally refer to the communication of values ​​and / or other data of syntax elements used to decode encoded video data. That is, video encoder 200 may signal the values ​​of syntax elements in the bitstream. Generally speaking, signaling refers to generating values ​​in the bitstream. As noted above, source device 102 may transmit the bitstream to destination device 116 substantially in real time or not in real time (such as when syntax elements are stored in storage device 112 for later retrieval by destination device 116).

[0095] According to the technology disclosed herein, video decoder 300 can receive video data indicating an image. Video decoder 300 can reconstruct blocks of video data to generate reconstructed blocks. Video decoder 300 can perform a neural network (NN)-based filtering process on the reconstructed blocks to generate filtered blocks. The NN-based filtering process may include applying an NN-based filter that includes a restricted header block. The restricted header block may have multiple input channels and is configured to extract a corresponding number of output channels for each of the multiple input channels and fuse the outputs associated with each of the multiple input channels into a single output. The sum of the corresponding number of output channels may be less than or equal to the number of fused output channels of the single output.

[0096] Video decoding standards include ITU-T H.261, ISO / IEC MPEG-1 Visual, ITU-T H.262 or ISO / IEC MPEG-2 Visual, ITU-T H.263, ISO / IEC MPEG-4 Visual, and ITU-T H.264 (also known as ISO / IEC MPEG-4 AVC), High Efficiency Video Decoding (HEVC), or ITU-T H.265, including its range extension, Multi-View Extension (MV-HEVC), and Scalable Extension (SHVC). Universal Video Decoding (VVC), or ITU-T H.266, has recently been developed by the Joint Video Experts Group (JVET) of the ITU-T Video Decoding Experts Group (VCEG) and the ISO / IEC Moving Picture Experts Group (MPEG).

[0097] The VVC specification has been finalized and is referred to below as VVC FDIS. It is available at phenix.int-evry.fr / jvet / doc_end_user / documents / 19_Teleconference / wg11 / JVET-S2001-v17.

[0098] Figure 2 This is a conceptual diagram illustrating a hybrid video decoding framework. Video decoding standards since H.261 are based on the so-called hybrid video decoding principle, which... Figure 2 The example is shown below. The term hybrid refers to a combination of two methods used to reduce redundancy in video signals: predictive and transform decoding with predictive residual quantization. Prediction and transform reduce redundancy in video signals by decorrelation, while quantization reduces the data represented by the transform coefficients by decreasing their precision, ideally by removing only irrelevant details. This hybrid video decoding design principle is also used in two recent standards, ITU-T H.265 / HEVC and ITU-T H.266 / VVC.

[0099] like Figure 2 As shown, a modern hybrid video decoder 130 typically performs block partitioning, motion compensation or inter-picture prediction, intra-picture prediction, transform, quantization, entropy decoding, and post-loop / intra-loop filtering. Figure 2 In the example, the video decoder 130 includes a summing 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 summing 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.

[0100] Generally, the video decoder 130 receives input video data 132 while encoding video data. Block partitioning is used to divide the received video data images into smaller blocks for use in the prediction and transformation processes. Early video decoding standards used a fixed block size, typically 16×16 samples. More recent standards such as HEVC and VVC employ tree-based partitioning structures to provide flexible partitioning.

[0101] Motion estimation unit 156 and inter-frame prediction unit 154 can predict input video data 132, for example, based on previously decoded data from DPB 150. Motion compensation, or inter-picture prediction, utilizes the redundancy present between pictures in the video sequence (hence the term "inter-picture"). According to 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 regions used to generate inter-frame predictions are indicated by motion information, including motion vectors and reference picture indices.

[0102] In recent video codecs, a hierarchical prediction structure within a group of pictures (GOP) is applied to improve decoding efficiency. Figure 3 An example hierarchical prediction structure 162 with a group of pictures (GOP) of size 16 is shown.

[0103] Refer again Figure 2 In-picture prediction utilizes the spatial redundancy present in the picture (hence the term "in-picture") by deriving predictions for blocks from spatially adjacent (reference) samples that have already been decoded / decoded. Angle prediction, DC prediction, and planar or on-plane prediction are used in recent video codecs, including AVC, HEVC, and VVC.

[0104] Hybrid video decoding standards can apply block transforms to prediction residuals (regardless of whether the prediction residuals originate from inter-image or intra-image predictions). Early standards, including H.261 / 262 / 263, employed Discrete Cosine Transform (DCT). In HEVC and VVC, more transform kernels beyond DCT can be applied to handle different statistical information in specific video signals.

[0105] Quantization aims to reduce the precision of input values ​​or a set of input values ​​in order to reduce the amount of data required to represent those values. In hybrid video decoding, quantization is typically applied to the residual samples after each transformation, such as to the transform coefficients, resulting in integer coefficient levels. In recent video decoding standards, the stride is derived from the so-called quantization parameter (QP), which controls fidelity and bit rate. A larger stride reduces the bit rate but also degrades quality, leading to issues such as block artifacts and blurred details in video images.

[0106] The summing unit 134 calculates the residual data as the difference between the input video data 132 and the predicted data from the intra-frame prediction unit 152 or the inter-frame prediction unit 154. The summing unit 134 provides the residual block to the transform unit 136, which applies one or more transforms to the residual block to generate a transform block. The quantization unit 138 quantizes the transform block to form quantized transform coefficients. The entropy decoding unit 140 entropy-encodes the quantized transform coefficients and other syntax elements (such as motion information or intra-frame prediction information) to generate an output bitstream 158.

[0107] Simultaneously, the inverse quantization unit 142 inverse quantizes the quantized transform coefficients, and the inverse transform unit 144 inverse transforms the transform coefficients to reconstruct the residual block. The summation unit 146 combines the residual block with the prediction block (on a sample-by-sample basis) to generate a decoded video data block. The 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 in-loop filter, or a predefined adaptive in-loop filter) to the decoded block to generate a filtered decoded block.

[0108] The entropy decoding unit 140 can perform context-adaptive binary arithmetic decoding (CABAC) on the encoded video. CABAC is used in recent video codecs such as AVC, HEVC, and VVC due to its high efficiency.

[0109] The loop filter unit 148 can perform post-loop filtering or intra-loop filtering. Post-loop / intra-loop filtering is a filtering process (or a combination of such processes) applied to reconstructed images to reduce decoding artifacts. The input to the filtering process is generally the reconstructed image, which is a combination of the reconstructed residual signal (including quantization errors) and the prediction. Figure 2 As shown, the reconstructed image after in-loop filtering is stored in the decoded image buffer (DPB) 150 and used as a reference for inter-image prediction of subsequent images. Decoding artifacts mainly depend on the QP. Therefore, QP information is often used in the design of the filtering process. In HEVC, the in-loop filter includes deblocking filtering and sample adaptive offset (SAO) filtering. In the VVC standard, an adaptive loop filter (ALF) is introduced as a third filter. The ALF filtering process is shown below:

[0110] (1)

[0111] in It can be a sample before the filtering process. These are the sample values ​​after the filtering process. Represents the filter coefficients. It is a clipping function, and This represents the clipping parameters. Variables k and l are... and The value varies between these values, where L represents the filter length. (Pruning function) Corresponding to function The pruning operation introduces non-linearity to make ALF more efficient by reducing the influence of neighboring sample values ​​that are very different from the current sample value. In VVC, the filtering parameters are signaled in the bitstream and can be selected from a predefined set of filters. The ALF filtering process can also be summarized by the following equation:

[0112] (2)

[0113] Figure 4 This is a conceptual diagram illustrating a neural network-based filter 170 with four layers. Various studies have shown that embedding neural networks into, for example... Figure 2 Compression efficiency can be improved in hybrid video decoding frameworks. Neural networks have been used for intra-frame and inter-frame prediction to improve prediction efficiency. In recent years, NN-based intra-loop filtering has also been an important research topic. In some examples, the filtering process is applied as a post-filter. In such examples, the filtering process is applied to the output image, and the unfiltered image can be used as a reference image.

[0114] In addition to existing filters such as deblocking filters, sample adaptive offset (SAO), and / or adaptive loop filtering (ALF), NN-based filters 170 can also be applied. NN-based filters can also be applied exclusively, where the NN-based filters are designed to replace all existing filters. Additionally or alternatively, NN-based filters (such as NN-based filter 170) can be designed to supplement, enhance, or replace any or all other filters.

[0115] Figure 4 The neural network-based filtering process can take reconstructed samples (e.g., luminance and chrominance samples, which in some examples can be packed into a 3D volume with 6 planes) as input, and the intermediate outputs are residual samples that are added back to the input to refine the input samples. The neural network-based filter can use all color components (e.g., Y, U, and V, or Y, Cb, and Cr, such as luminance data 172A, blue hue chrominance 172B, and red hue chrominance 172C) as input to take advantage of cross-component correlations. Different color components can share the same filters (including network structure and model parameters), or each component can have its own specific filters.

[0116] The filtering process can also be summarized as follows:

[0117] (3)

[0118] The model structure and parameters of the NN-based filter can be predefined and stored in the video encoder 200 and video decoder 300. The filter can also be notified by signals sent in the bitstream.

[0119] exist Figure 4 In the example, a neural network-based filter could include a series of feature extraction layers followed by an output convolution. Figure 4 In a convolutional layer, the feature extraction layer may include a 3×3 convolutional (conv) layer followed by a Parametric Rectified Linear Unit (PReLU) layer. The convolutional layer applies convolution operations to the input data, which involves a filter or kernel sliding across the input data (e.g., a reconstructed sample of input 172) and computing the dot product at each location. The convolutional operation essentially captures local patterns within the input data. For example, in the context of image processing, these patterns could be edges, textures, or other visual features. The filter or kernel is a small matrix of weights that is updated during the training process. By sliding the filter across the input data (or a feature map from the previous layer) and computing the dot product at each location, the convolutional layer creates a feature map that encodes the spatial hierarchy and patterns detected in the input.

[0120] The output of a convolutional layer is a set of feature maps, each corresponding to a filter, thus capturing different aspects of the input data. As data passes through deeper layers of the network, this layer helps the neural network learn increasingly complex and abstract features. Figure 4 In the naming convention 3×3 conv 3×3×6×8, the first 3×3 indicates that the convolutional layer has a 3×3 filter size (e.g., a 3×3 matrix). 3×3×6×8 refers to both the input and output dimensions of the convolutional layer, where 6 is the number of input channels and 8 is the number of output channels.

[0121] The PReLU layer is an activation function used in neural networks and is introduced as a variant of the ReLU (Rectified Linear Unit) activation function. As described above, convolutional layers output feature maps, each corresponding to a filter used to represent the detected features in the input. After the convolutional layers, the PReLU layer applies the PReLU activation function to each element of the feature map produced by that convolutional layer. For positive values, the PReLU layer acts similarly to standard ReLU, allowing the value to pass through. For negative values, the PReLU layer allows small linear negative outputs instead of setting them to zero (e.g., as ReLU does). This keeps neurons active and maintains gradient flow, which can be beneficial for learning in deep networks.

[0122] 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 non-linearity and helps avoid neuron death by allowing small gradients when the input is negative. This combination is effective when learning complex patterns in the data while maintaining a robust gradient flow, which is particularly beneficial in deeper network architectures.

[0123] When applying neural network-based filtering in video decoding, the entire video signal (pixel data) may be divided into multiple processing units (e.g., 2D blocks), and each processing unit may be processed individually or combined with other information associated with that pixel block. Possible choices for processing units include frames, slices / tiles, CTUs, or any predefined or signaled shape and size. Typically, neural network-based filtering is performed on reconstructed blocks of video data. Here, reconstructed blocks and samples may refer to decoded blocks generated by video decoder 300 and blocks reconstructed in the reconstruction loop of video encoder 200.

[0124] To further improve the performance of neural network-based filtering, different types of input data can be jointly processed to produce filtered outputs. Input data may include, but is not limited to, reconstructed pixels / samples, predicted pixels / samples, pixels / samples after loop filtering, partitioning structure information, deblocking parameters (e.g., boundary strength (BS)), QP values, slice or image type, or filter suitability or decoding mode diagrams. Input data can be provided at different granularities. Luminosity reconstruction and predicted samples can be provided at the original resolution, while chromosity samples can be provided at a lower resolution (e.g., for 4:2:0 representation), or can be upsampled to luminosity resolution for per-pixel representation. Similarly, QP, BS, partitioning, or decoding mode information can be provided at a lower resolution, including cases where each frame, slice, or processing block (e.g., QP) has a single value. In other examples, QP, BS, partitioning, or decoding mode information can be expanded (e.g., replicated) for per-pixel / sample representation.

[0125] Figure 5 An example of an architecture utilizing supplementary data is shown in the figure. Figure 5This is a block diagram illustrating an example CNN-based filter with padded input samples and supplementary data. The NN-based filter 171 uses pixels / samples of a processing block combined with supplementary data as input 174. Input 174 may include four sub-blocks 174A of interleaved luminance samples (Y×4) and associated blue hue chromaticity (U) data 174B and red hue chromaticity (V) data 174C. Supplementary data includes quantization parameter (QP) step size 176 and boundary strength (BS) 178. The region of input pixels / samples can be expanded from each side using four padded pixels / samples. The resulting size of the processing volume is (4+64+4)×(4+64+4)×(4Y+2UV+1QP+3BS).

[0126] Compared to Figure 4 The neural network-based filter 171 may include two or more hidden layers utilizing both 1×1 convolutions and Leaky ReLU layers. The Leaky ReLU layer, similar to the PReLU layer, allows the output of a small non-zero gradient when the layer is inactive. Instead of outputting zero for negative inputs, Leaky ReLU multiplies these inputs by a small constant. This small slope ensures that even neurons that would otherwise be inactive still contribute a small amount to the network's learning, thus reducing the likelihood of ReLU dying.

[0127] To further improve the performance of neural network-based filtering, multi-mode solutions 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 can signal this selection in the bitstream. Different modes may include different neural network models, different values ​​of input information used as these neural network models, etc. In one example, the video encoder 200 and the video decoder 300 can use a multi-mode neural network-based filtering solution based on a single neural network model by using different QP values ​​as inputs to the neural network model for different modes.

[0128] An example of an In-Loop Filter (ILF) architecture being developed by JVET is described below. One such architecture is a multimodal CNN ILF with attention blocks. A multimodal NN-based filtering solution is disclosed for NN-based filtering with multimodal design in Y. Li, K. Zhang, L. Zhang, H. Wang, M. Coban, AMKotra, M. Karczewicz, F. Galpin, K. Andersson, J. Ström, D. Liu, R. Sjöberg, April 2022, EE1-1.7: Combined Test of EE1-1.6 and EE1-1.3, JVET-Z0113 (hereinafter referred to as "JVET-Z0113"). This solution is described above.

[0129] In one example, a neural network-based filtering solution with multiple modes can be used, as described above. Figure 6 The structure of a neural network is shown in the diagram. Figure 6 In the first part, different input data types are convolved with multiple kernels of size 3×3 (conv3×3) to generate feature maps. The convolved inputs are activated by parametric rectified linear units (PReLU), and then the results of each data type are concatenated (merged), fused, and subsampled once (transformed) to create the output y. This output is then fed through attention residual (AttRes) blocks, each attention residual block having Figure 6 The structure shown is used to feed the output from the last attention residual block z into the final part of the network.

[0130] For example, Figure 6 The NN-based filter includes a first part comprising input 3×3 convolutions 510A to 510E for each input in the input and corresponding parametric rectified linear units (PReLU) 512A to 512E to generate feature maps (e.g., feature extraction segments of the NN-based filter). A cascade unit 514 cascades the feature maps and provides them to a fusion block 516 and a transformation block 522. Figure 6 The NN-based filter also includes a set 528 of attention residual (AttRes) blocks 530A to 530N; and a final portion (e.g., a tail segment) including 3×3 convolutions 550, PReLU 552, 3×3 convolutions 554, and pixel recombination units 556. The AttRes blocks may also be referred to as backbone blocks.

[0131] In the first part (e.g., the feature extraction section), different inputs are received, including quantization parameters (QP) 500, partitioning information (part) 502, boundary strength (BS) 504, predicted samples (pred) 506, and reconstructed samples (rec) 508. Corresponding 3×3 convolutions 510A to 510E and PReLU 512A to 512E convolve the corresponding inputs and activate them to generate feature maps. Then, cascade unit 514 concatenates these feature maps. A fusion block 516, including 1×1 convolutions 518 and PReLU 520, fuses the concatenated feature maps. A transform block, including 3×3 convolutions 524 and PReLU 526, subsamples the fused inputs to create output 188. Output 188 is then fed through a set 528 of attention residual blocks 530A to 530N, which may include a variety of numbers of attention residual blocks, such as eight. This attention block is relative to... Figure 7 Further explanation. The output 189 of the last attention residual block from the set 528 of attention residual blocks 530 is fed into the final part of the NN-based filter. In the final part (which may be the tail block), 3×3 convolutions 550, PReLU 552, 3×3 convolutions 554, and a pixel reconstruction unit 556 process the output 189, and an adder unit 558 combines the result with the original input reconstructed sample 508. This ultimately forms a filtered output for presentation and storage (e.g., in a decoded image buffer (DPB)) as a reference for subsequent inter-frame prediction. In some examples, Figure 6 The NN-based filter uses 96 feature maps.

[0132] Figure 7 This is an example Figure 6 A conceptual diagram of the attention residual block. That is, Figure 7 An attention residual block 530 is depicted, which may include similar attention residual blocks. Figure 6 The components are components of attention residual (AttRes) blocks 530A to 530N. In this example, attention residual block 530 includes a first 3×3 convolution 532, a parametric rectified linear unit (PReLU) filter 534, a second 3×3 convolution 536, attention block 538, and an adder unit 540. Adder unit 540 combines the output of attention block 538 with the output 188 initially received by the first 3×3 convolution 532 to generate output 189.

[0133] Figure 8 This is an example Figure 7 A conceptual diagram of an example spatial attention layer. (See also:) Figure 8As shown, the spatial attention layer of the attention residual block 530 includes a 3×3 convolution 706, a PReLU 708, a 3×3 convolution 710, a size expansion unit 712, a 3×3 convolution 720, a PReLU 722, and a 3×3 convolution 724. The 3×3 convolution 706 receives input 702, which corresponds to... Figure 6 The quantization parameters (QP) are 500, the partition information (part) is 502, the boundary strength (BS) is 504, the predicted sample (pred) is 506, and the reconstructed sample (rec) is 508. A 3×3 convolution with a 720-bit reception Z... K Value 704. The output of the size expansion unit 712 is combined with the output of the 3×3 convolution 724, 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 Values ​​704 are combined to generate output Z. K+1 Value 734.

[0134] Another example neural network architecture is the reduced-complexity multimodal CNN ILF. An alternative design for the neural network architecture is proposed in S. Eadie, M. Coban, M. Karczewicz's January 2023 paper, EE1-1.9: Reduced complexity CNN-based in-loop filtering, JVET-AC0155 (hereinafter referred to as "JVET-AC0155"). For example, one could use... Figure 6 The example filter incorporates a large number of low-complexity residual blocks in its backbone, reduces the number of channels (feature maps), and removes the attention module. The proposed CNN-based filtering structure (e.g., for brightness filtering) achieves [the desired effect]. Figure 9 As shown in the image.

[0135] Figure 9 This is a block diagram illustrating a simplified CNN-based filter architecture with padded input samples and supplementary data. Figure 9 The NN-based filters include 3×3 convolutions 810A to 810E and PReLU 812A to 812E, which convolve the corresponding inputs (i.e., QP 800, Part 802, BS 804, Pred 806, and Rec 808) to generate feature maps (e.g., feature extraction segments). Concatenation unit 814 concatenates the convolved inputs (e.g., feature maps). Fusion block 816 then uses 1×1 convolutions 818 and PReLU 820 to fuse the concatenated feature maps. Transformation block 822 then uses 3×3 convolutions 824 and PReLU 826 to process the fused data.

[0136] In this example, the NN-based filter comprises a set 828 of residual blocks 830A to 830N (also referred to as backbone blocks), each of which can be configured according to the following discussion. Figure 10 The residual block structure 830 is used for construction. Residual blocks 830A to 830N are replaceable. Figure 6 AttRes blocks 530A to 530N. Figure 9 The example can be used for luminance (luminance) filtering, but similar modifications can be made for chrominance (chrominance) filtering as discussed below.

[0137] Figure 9 The number of residual blocks and channels included in set 828 can be configured in different ways. That is, N can be set to different values, and the number of channels in 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 to process chroma channels.

[0138] The set 828 of residual blocks 830A to 830N has N instances of residual block structure 830. In one example, N can be equal to 32, such that there are 32 residual block structures. Residual blocks 830A to 830N can use 64 feature maps, which are relative to... Figure 6 The 96 feature maps used in the example have been reduced.

[0139] exist Figure 9 In the final part, 3×3 convolution 850, PReLU 852, 3×3 convolution 854, and pixel reconstruction unit 856 process the output of set 828, and addition unit 858 combines the result with the original input reconstruction sample (REC) 808. This ultimately forms a filtered output for presentation and storage (e.g., in a decoded image buffer (DPB)) as a reference for subsequent inter-frame prediction.

[0140] In one example, Figure 9 In the example, the number of residual blocks and feature maps (convolutions) used are M=24 and 64, respectively. In residual blocks 830A to 830N, the number of channels first increases to 160 before the activation layers and then decreases to 64 after the activation layers. For different performance-complexity tradeoffs, the number of residual blocks and channels can be configured differently (M is set to another value and the number of channels in the residual blocks can be set to a number different from 160). Chroma filtering can follow... Figure 6 The concept involves modifying its core as described above to handle chroma channels.

[0141] In yet another NN architecture, residual blocks can be Figure 11The filter block replacement shown is discussed later in this disclosure. In this example, the bypass branches around the convolutional and activation layers in the residual blocks of the previous solution are removed. The number of channels and the number of filter blocks can be configured, for example, 64 channels and 24 filter blocks, with 160 channels before and after activation, resulting in a network complexity of 605.93kMAC (kMAC is one thousand multiply-accumulate operations) and 1.5M parameters for the intra-frame luma model.

[0142] Another example neural network architecture is the multimodal CNN ILF that utilizes separable convolutions. Replacing 2D convolutions (3×3) with separable convolutions further reduces the complexity of the CCN ILF structure. In the April 2023 EE2 report by V. Seregin, J. Chen, R. Chernyak, K. Naser, J. Ström, F. Wang, M. Winken, X. Xiu, and K. Zhang (EE1 Test 1.3.5), the low-rank convolution approximation decomposes a 3×3×M×N convolution into a pixel-wise convolution (1×1×M×R), two separable convolutions (3×1×R×R, 1×3×R×R) (e.g., two convolutions separated or divided into separate convolutions of the original convolution) and another pixel-wise convolution (1×1×R×N) are residual blocks applied to the architecture described in JVET-AC0155. In this example, R is the approximate rank and the approximate performance / complexity can be cut off.

[0143] Figure 10 This is an example Figure 9 A conceptual diagram of an example residual block structure 830 is provided. In this example, the residual block structure 830 includes a first 1×1 convolution 832, which increases the number of input channels to 160 before processing the input channels in the activation layer (PReLU 834). PReLU 834 can thus reduce the number of channels to 64 through this processing. A second 1×1 convolution 836 then processes the reduced channels, followed by a 3×3 convolution 838. Finally, a combining unit 840 combines the output of the 3×3 convolution 838 with the original input received by the residual block structure 830.

[0144] In yet another NN architecture, residual blocks can be... Figure 11 The filter block (also known as the backbone block) is shown as being replaced. In this example, bypass branches around all layers in each residual block are removed, as shown... Figure 12 As shown. The number of channels and filter blocks is configurable, for example, 64 channels and 24 filter blocks, with 160 channels before and after activation. In some examples, the complexity of such a network is 605.93kMAC and the number of parameters used for the internal luminance model is 1.5M.

[0145] Figure 11 This illustrates alternatives to the technology according to this disclosure. Figure 6 A conceptual diagram of another example filter block structure of an attention residual block set. Figure 11 The NN-based filters include 3×3 convolutions 1010A to 1010E and PReLU 1012A to 1012E, which convolve the corresponding inputs (i.e., QP 1000, Part 1002, BS 1004, Pred 1006, and Rec 1008) to form feature maps (e.g., feature extraction segments). Concatenation unit 1014 concatenates these feature maps. Fusion block 1016 then uses 1×1 convolution 1018 and PReLU 1020 to fuse the concatenated inputs. Transformation block 1022 then uses 3×3 convolution 1024 and PReLU 1026 to process the fused data.

[0146] In this example, the NN-based filtering unit comprises a set 1028 of N filter blocks 1030A to 1030N (also referred to as the backbone block), each filter in the filter block may have Figure 12 The structure of filter block 1030 is discussed below. Filter block structure 1030 is substantially similar to residual block structure 830, except that the combination unit 840 is omitted from filter block structure 1030, so that the input is not combined with the output. Instead, the output of each residual block structure can be directly fed to subsequent blocks.

[0147] exist Figure 11 In the final part, 3×3 convolution 1050, PReLU 1052, 3×3 convolution 1054, and pixel reconstruction unit 1056 process the output of set 1028, and addition unit 1058 combines the result with the original input reconstructed sample (REC) 1008. This ultimately forms a filtered output for presentation and storage (e.g., in a decoded image buffer (DPB)) as a reference for subsequent inter-frame prediction.

[0148] In one example Figure 11 The architecture is implemented using parameters K=64, M=160 and R=51, and the total number of residual blocks of 24 makes the network complexity 356.43kMAC, and the number of parameters used for the intra-frame luminance model is 1.07M.

[0149] Figure 12 This is an example Figure 11 A conceptual diagram of an example filter block structure 1030 is provided. In this example, the filter block structure 1030 includes a first 1×1 convolution 1032, which increases the number of input channels to 160 before processing the input channels in the activation layer (PReLU 1034). PReLU 1034 can thus reduce the number of channels to 64 through this processing. A second 1×1 convolution 1036 then processes the reduced channels, followed by a 3×3 convolution 1038. As discussed above, with... Figure 10 Compared to the residual block structure 830, the filter block structure 1030 does not include a combination unit.

[0150] Convolutions with 3×3 kernels are popular in neural network-based filters. In the architecture described above, 3×3×N×M convolutions are utilized across multiple sections and blocks, where the 3×3 kernels slide in the spatial (2D) domain. However, multidimensional convolutions such as 2D kernel convolutions introduce significant complexity. The video encoder 200 and video decoder 300 can be configured to utilize separable convolutions to replace or supersede multidimensional convolutions (e.g., 3×3×N×M convolutions). For example, two separable (e.g., separate) one-dimensional convolutions can be used to replace 3×3 convolutions in any segment of a neural network-based filter. The use of separable convolutions reduces computational complexity and memory bandwidth requirements.

[0151] To avoid overcomputation and reduce the parameter set derived from multidimensional convolutions (such as 3×3 convolutions (or 2D convolutional quantities with higher-dimensional kernels) or similar convolutions in the CNN architecture described above), this disclosure describes a technique in which the video encoder 200 and video decoder 300 are configured to utilize separable convolutions (e.g., 1D separable convolutions) generated by low-complexity approximations instead of multidimensional (e.g., 2D) convolutions that slide in the spatial direction. While the technique of this disclosure is described with reference to 3×3 convolutions, the decomposition technique of this disclosure can be used for multidimensional convolutions of any size (e.g., 4×4, 5×5, or larger). In general, a multidimensional convolution has a kernel size n1×n2 in the spatial dimension, where n1 and n2 are positive integers. The values ​​of n1 and n2 can be the same or different. A multidimensional convolution can also have a size K (e.g., n1×n2×K) in the depth dimension. Furthermore, when the number of output channels is M, a multidimensional convolution can be represented as a 4-D tensor of n1×n2×K×M.

[0152] In one example of this disclosure, the low-rank convolution approximation decomposes a 3×3×M×N convolution into a pixel-wise convolution (1×1×M×R), two separable convolutions (3×1×R×R, 1×3×R×R), and another pixel-wise convolution (1×1×R×N). 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 to be 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.

[0153] In a typical example, a multidimensional convolution can be approximated by multiple separable convolutions in the following way: perform a first convolution of size n1×1 and perform a second convolution of size 1×n2 on the output of the first convolution.

[0154] Figure 13 This is a block diagram illustrating an example multi-scale feature extraction backbone network with two-component convolution. Figure 13 Examples can be similar to Figure 10 The example uses a 3×3×K×K convolution, which is an approximation of a 3×1×K×R convolution and a 1×3×R×K convolution.

[0155] exist Figure 13 In the example, residual block 1420 comprises a 1×1×K×M convolution 1402, followed by a PReLU 1404. The output of PReLU 1404 is fed into a 1×1×M×K convolution 1406. The 3×3×K×K convolution 1408 of residual block 1420 is approximated by a 3×1×K×R convolution 1400, and then by a 1×3×R×K convolution 1410. The output of the 1×3×R×K convolution 1410 can be fed into a combining unit 1412, which combines the output of the 1×3×R×K convolution 1410 with the input of the 1×1×K×M convolution 1402. R is the canonical rank of this decomposition. A lower rank implies a greater reduction in complexity.

[0156] Figure 14 This is a block diagram illustrating an example unified filter with a joint model (joint luminance and chrominance). It is presented in Y. Li, S. Eadie, D. Rusanovskyy, M. Karczewicz, April 2023, EE1 Related: Combinatorial Testing of Multiscale Components of EE1-1.3.5 and EE1-1.6, JVET-AD0211 (hereinafter referred to as "JVET-AD0211"). Figure 14The illustrated example utilizes a two-component convolutional network for multi-scale feature extraction. A 3×3 convolution can be decomposed into a 3×1×C1×R convolution followed by a 1×3×R×C2 convolution, where C1 and C2 are the number of input and output channels, respectively, and R is the approximate rank (see V. Lebedev, Y. Ganin, M. Rakhuba, I. Oseledets, V. Lempitsky, Speeding up Convolution Neural Networks Using Fine-tuned CP-Decomposition, ICLR 2015, available online: https: / / arxiv.org / pdf / 1412.6553). In some examples, the parameter R can be derived proportionally to R = C1xC2 / (C1+C2) and control the approximate complexity.

[0157] Figure 14 The proposed architecture is shown, where 3×3 convolutional blocks are replaced by separable convolutions of 3×1 and 1×3. In this example, the residual block structure 1430 includes a first 1×1 convolution 1432 before the first activation layer (PReLU 1434), and a 3×3 convolution 1440 and a second activation layer (PReLU 1442) parallel to the first 1×1 convolution 1432 and PReLU 1434. The second 1×1 convolution 1436 then processes the combined output of PReLU 1434 and PReLU 1442, followed by a 3×3 convolution 1438. However, in Figure 14 In the example, a 3×3 convolution 1440 can be approximated using multiple separable convolutions, such as... Figure 14 The 3×1 convolution 1450 and 1×3 convolution 1452 are shown in the figure. Similarly, the 3×3 convolution 1438 can be approximated by multiple separable convolutions, such as... Figure 14 The 3×1 convolution 1460 and 1×3 convolution 1462 are shown in the figure.

[0158] As an example, Figure 14 The illustrated architecture can be implemented using parameters R1=8, R2=44, M1=160 and M2=16, and the total number of residual blocks of 24 makes the network complexity 358.43kMAC, and the number of parameters used for the intra-frame luminance model is 1.07M.

[0159] Figure 15 This is a block diagram illustrating an example unified filter with a joint model (joint luminance and chrominance). Figure 15 NN-based filters may include IPB 1500 (which provides information about whether a block is inter-frame prediction or intra-frame prediction), QP SLICE1502 (which provides quantization parameters for slicing), QP BASE 1504 (which provides quantization parameters for the sequence), Boundary Strength (BS) 1506, Predicted Samples (PRED) 1508, and Reconstructed Samples (REC) 1509. The video decoder 300 can apply the corresponding 3×3 convolutions from 1510A to 1510F to each input in the input. The video decoder 300 can apply the corresponding PReLU from 1512A to 1512F to the output of the 3×3 convolutions from 1510A to 1510F. In some examples, backbone blocks 1530A to 1530N can be residual blocks.

[0160] The fusion block 1516 uses a 1×1 convolution 1518 and PReLU 1520 to fuse the feature maps. The transformation block 1522 then uses a 3×3 convolution 1524 and PReLU 1526 to process the fused data.

[0161] The video decoder 300 may apply a set 1528 of backbone blocks. Set 1528 may include multiple filter blocks 1530A to 1530N (also referred to herein as backbone blocks). In some examples, N=24, resulting in 24 filter blocks. The video decoder 300 may apply a 3×3 convolution 1550 to the output of set 1528. The video decoder 300 may apply a PReLU 1552 to the output of the 3×3 convolution 1550. The video decoder 300 may apply a 3×3 convolution 1554 to the output of the PReLU 1552. The video decoder 300 may crop 1558 of the output of the 3×3 convolution 1554 to generate filtered reconstructed UV (chroma) samples of the image. The video decoder 300 may perform pixel reconstruction 1556 on the output of the 3×3 convolution 1554 and crop 1560 pixels of the reconstructed 1556 output to generate filtered reconstructed Y samples of the image. For example, pixel recombination 1556 can upsample its input so that the output of pixel recombination 1556 has w 2, h 2, c / 4, where w is the width, h is the input height, and c is the number of input channels to the 1556 pixel reassembly input. In some examples, Figure 15 The example parameters include one or more of the following: d1 = 192, d2 = 32, d3 = 16, d4 = 16, d5 = 16, C = 64, and d6 = 48.

[0162] Figure 16 This is a block diagram illustrating the example main block. Figure 16The backbone block 1600 can be an example of any filter block from 1530A to 1530N. The backbone block 1600 may include multi-scale branches 1620, where convolutions are performed in parallel. For example, a 1×1 convolution 1602 may be performed in parallel with a 3×1 convolution 1604, followed by a 1×3 convolution 1606. The multi-scale branches 1620 may perform multi-scale feature extraction.

[0163] Video decoder 300 can apply PReLU 1608 to the multi-scale branch 1620 of the output. Video decoder 300 can apply a 1×1 convolution 1610 to the output of PReLU 1608. Video decoder 300 can apply a 1×3 convolution 1612 to the output of 1×1 convolution 1610. Video decoder 300 can apply a 3×1 convolution 1614 to the output of 1×3 convolution 1612. In some examples, Figure 16 The example parameters include C = 64, C1 = 160, C21 = 32, C22 = 32, and C31 = 64.

[0164] Figure 17 This is a block diagram illustrating an example unified filter with separate luminance and chrominance models (luminance). Figure 17 NN-based filters may include IPB 1700, QP BASE 1702, BS 1704, PRED 1706 and reconstructed luminance samples (REC) EXT Y)1708. The video decoder 300 can apply the corresponding 3×3 convolutions from 1710A to 1710E to each input in the input. The video decoder 300 can apply the corresponding PReLUs from 1712A to 1712E to the output of the 3×3 convolutions from 1710A to 1710E.

[0165] The fusion block 1716 uses a 1×1 convolution 1718 and PReLU 1720 to fuse the feature maps. The transformation block 1722 then uses a 3×3 convolution 1724 and PReLU 1726 to process the fused data.

[0166] The video decoder 300 may apply a set 1728 of backbone blocks. Set 1728 may include multiple filter blocks 1730A to 1730N (also referred to herein as backbone blocks). In some examples, N=20, resulting in 20 filter blocks. The video decoder 300 may apply a 3×3 convolution 1750 to the output of set 1728. The video decoder 300 may apply a PReLU 1752 to the output of the 3×3 convolution 1750. The video decoder 300 may apply a 3×3 convolution 1754 to the output of the PReLU 1752. The video decoder 300 may crop 1758 of the output of the 3×3 convolution 1754 to generate filtered reconstructed UV samples of the image. The video decoder 300 may perform pixel reconstruction 1756 on the output of the 3×3 convolution 1754 and crop 1760 pixels of the reconstructed 1756 output to generate filtered reconstructed Y samples of the image.

[0167] One or more filter blocks from 1730A to 1730N in the set 1728 can be Figure 16 An example of a 1600-byte backbone block. In some examples, Figure 17 Examples along with Figure 16 The parameters for the example may include one or more of the following: d1 = 192, d2 = 32, d3 = 16, d4 = 16, d5 = 16, C = 64, C1 = 160, C 21 = 32、C 22 =32、C 31 = 64, N = 20 and d6 = 48.

[0168] Figure 18 This is a block diagram illustrating an example unified filter with separate luminance and chrominance models (chrominance). Figure 18 The input to the NN-based filter may include REC EXT Y 1800 (which can be downsampled by 2 times), QP BASE 1804, BS 1806, PRED1808 and reconstructed chromaticity samples (REC) EXT (UV) 1809. The video decoder 300 can apply the corresponding 3×3 convolution from 1810A to 1810E to each input in the input. The video decoder 300 can apply the corresponding PReLU from 1812A to 1812E to the output of the 3×3 convolution from 1810A to 1810E.

[0169] The fusion block 1816 uses a 1×1 convolution 1818 and PReLU 1820 to fuse the feature maps. The transformation block 1822 then uses a 3×3 convolution 1824 and PReLU 1826 to process the fused data.

[0170] The video decoder 300 may apply a set 1828 of backbone blocks. Set 1828 may include multiple filter blocks 1830A to 1830N (also referred to herein as backbone blocks). In some examples, N=16, resulting in 24 filter blocks. The video decoder 300 may apply a 3×3 convolution 1850 to the output of set 1828. The video decoder 300 may apply a PReLU 1852 to the output of the 3×3 convolution 1850. The video decoder 300 may apply a 3×3 convolution 1854 to the output of the PReLU 1852. The video decoder 300 may crop 1858 of the output of the 3×3 convolution 1854 to generate filtered reconstructed UV samples of the image. The video decoder 300 may perform pixel reconstruction 1856 on the output of the 3×3 convolution 1854 and crop 1858 of the output of the pixel reconstruction 1856 to generate filtered reconstructed Y samples of the image.

[0171] One or more filter blocks from 1830A to 1830N in the 1828 filter block set can be used Figure 16 The backbone architecture. In some examples, Figure 18 Together Figure 16 The parameters for the example may include one or more of the following: d1 = 192, d2 = 32, d3 = 16, d4 = 16, d5 = 16, C = 64, C1 = 160, C 21 = 32、C 22 = 32、C 31 = 64, N = 16 and d6 = 48.

[0172] A multi-scale feature extraction backbone with two-component decomposition has been integrated into the unified model in JVET's Exploratory Experiment (EE). Additionally, the specifications from the EE include two versions of the model: 1) a unified model combining luminance and chrominance, such as... Figure 15 As shown; and 2) separate models for luminance and chromaticity, respectively as shown in... Figure 17 and Figure 18 As shown.

[0173] The unified filter (UF) architecture used by JVET (such as Figures 15 to 18 (As shown) For video decoding applications at relatively high performance layers, operation is performed at speeds exceeding 400kMAC / pixel to decode YCbCr 4:2:0 data. To run in-loop filters with video codecs on state-of-the-art computationally and / or power-constrained devices, the complexity of NN-based tools is constrained to below 20kMAC / pixel. Therefore, implementation is not feasible on such computationally and / or power-constrained devices. Figures 15 to 18 The UF architecture.

[0174] To achieve a complexity level below 20k MAC / pixel while retaining most of the decoding performance, an alternative UF architecture is disclosed in Y. Li, D. Rusanovskyy, M. Karczewicz's July 2023 paper EE1-4.4, "Low complexity NN filter with design elements of Unified Filter Architecture," and in EE1-1.2 and EE1-1.3, JVET-AE0165. This alternative architecture replaces the 3×3 convolutions in the backbone network with regularized clustering (CP) decomposition featuring separable convolutions. Figure 19 As shown.

[0175] Figure 19 This is a block diagram illustrating an example low-complexity in-loop filter network architecture according to one or more aspects of this disclosure. For example, Figure 19 The architecture can be used to implement low-complexity in-loop filters for joint YCbCr 4:2:0 processing. Figure 19 The example also applies to the separate component Y,CbCr processing. Figure 19 An example could be a uniform filter. A uniform filter could include a filter with at least one quality input (such as a QP input), making the filter usable for handling different quality levels. Figure 19 It can be seen that, Figure 19 The example filter is one with QP SLICE 1902 Input and QP BASE 1904. A unified filter is input to both. A variety of image and / or video data of different quality levels can be used to train the NN-based unified filter (such as...). Figure 19 The filter shown.

[0176] Figure 19 The architecture includes a header block 1980, a transition block 1922, a set 1928 of one or more backbone blocks (e.g., filter blocks 1930A to 1930N), and a tail block 1986. The inputs of the header block 1980 include IPB 1900 and QP... SLICE 1902, QP BASE 1904, BS 1906, PRED 1908 and REC 1909.

[0177] Head block 1980 includes blocks preceding fusion block 1916. Figure 19 The filter block. Header block 1980 includes 3×3 convolutions 1910A to 1910F applied to the corresponding inputs.

[0178] Fusion block 1916 can use 1×1 convolution 1918 and PReLU 1920 to fuse feature maps. Transformation block 1922 may include elements appearing after fusion block 1916 and before any backbone block or filter block (e.g., before filter blocks 1930A to 1930N). Figure 19 The filter blocks. Transformation block 1922 uses 3×3 convolution 1924 and PReLU 1926 to process the fused data. Set 1928 may include filter blocks 1930A to 1930N (also referred to as backbone blocks). In some examples, filter blocks 1930A to 1930N provide Figure 19 The filtering function of the filter. In some examples, filter blocks 1930A to 1930N may be located after transition block 1922 and before tail block 1986.

[0179] Tail block 1986 may include blocks located after all trunk blocks or filter blocks (e.g., after filter blocks 1930A to 1930N). Figure 19 The filter block. The tail block 1986 may include a 3×3 convolution 1950, a PReLU 1952, a 3×3 convolution 1954, a cropping unit 1958, a pixel recombination unit 1956, and a cropping unit 1960.

[0180] Figure 19 Examples of sample architectures and parameters are shown, such as the number of inputs, features, channels, and the number of residual blocks. In some examples, Figure 19 The example parameters include one or more of the following: d1 = 12, d2 = 8, d3 = 4, d4 = 2, d5 = 2, C = 24, N = 11, and d6 = 24. For the illustrated configuration, the complexity is approximately 19.55kMAC.

[0181] Figure 20 This is a block diagram illustrating an example backbone block according to one or more aspects of this disclosure. Backbone block 2000 may be... Figure 19 Examples of any filter blocks from filter blocks 1930A to 1930N in set 1928. The backbone block 2000 differs from the UF backbone block 1600 in that backbone block 2000 does not include multi-scale branches as in backbone block 1600, and the multidimensional convolutions replaced by convolutions 1610, 1612, and 1614 are replaced by CP decomposition in backbone block 2000. Backbone block 2000 includes 1×1 convolution 2002 and PReLU 2004. The CP decomposition in backbone block 2000 includes 1×1 convolution 2006, 1×3 convolution 2008, 3×1 convolution 2010, and 1×1 convolution 2012. In some examples, Figure 20 The parameters include C = 24, C1 = 72 and C21 = 24.

[0182] Figure 21 This is a block diagram illustrating an example CP decomposition according to one or more aspects of this disclosure. CP decomposition 2100 may include a 1×1 convolution 2106, a 1×3 convolution 2108, a 3×1 convolution 2110, and a 1×1 convolution 2112. For example, CP decomposition 2100 may replace the 3×3 convolution 1950 of the tail block 1986. In some examples, Figure 21 The parameters include C = 24, C1 = 72 and C21 = 24.

[0183] Figure 22 This is a block diagram illustrating an example low-complexity in-loop filter network architecture with luminance / chrominance separation. Figure 22 This refers to the low-complexity architecture described in HG9, July 2023, "Common text for proposed generative face video SEI message" by B. Chen, J. Chen, Y. Ye, S. Wang, S. McCarthy, P. Yin, G.-M. Su, AKChoudhury, and W. Husak, JVET-AE0281, where the processing branches are split into separate luma and chroma processing branches, which allows for a better balance of luma / chroma performance.

[0184] Figure 22 The architecture includes a header block 2280, a fusion block 2222, one or more chroma backbone blocks 2228, one or more luma and chroma backbone blocks 2230, a tail block 2286, and a tail block 2288. The inputs to the header block 2280 include IPB 2200 and QP... SLICE 2202, QP BASE 2204, BS 2206, PRED 2208 and REC 2209.

[0185] Head block 2280 is included before cascade 2214 Figure 22 The filter block. The header block 2280 includes 1×1 convolutions 2210A to 2210D and 3×3 convolutions 2210E to 2210F applied to the corresponding input.

[0186] The fusion block 2222 can use a 1×1 convolution 2218 and PReLU 2220 to fuse feature maps. The fusion block 2222 can also process the output of PReLU 2220 by applying a 3×3 convolution 2224 and PReLU 2226. One or more chroma backbone blocks 2228 can provide chroma samples. Figure 22The filtering function of the filter. One or more luminance and chrominance backbone blocks 2230 can provide luminance and chrominance samples. Figure 22 The filtering function of the filter.

[0187] Tail block 2286 may include blocks located after the backbone blocks (e.g., one or more chroma backbone blocks 2228 and one or more luminance and chroma backbone blocks 2230). Figure 22 The tail block 2286 may include a separable convolution 2250, which may include 1×3 and 3×1 convolutions, PReLU 2252 and 3×3 convolutions 2254. The tail block 2288 may include a separable convolution 2260, which may include 1×3 and 3×1 convolutions, PReLU 2262 and 3×3 convolutions 2264. The output of the tail block 2288 may be input to the pixel recombination unit 2266.

[0188] Figure 23 This is an example backbone block according to one or more aspects of this disclosure. Backbone block 2300 may be... Figure 22 Examples of backbone blocks 2300 include one or more chroma backbone blocks 2228 and / or one or more luma and chroma backbone blocks 2230. Backbone block 2300 includes a 1×1 convolution 2302, a PReLU 2304, a 1×1 convolution 2306, a 1×3 separable convolution 2308, a 3×1 separable convolution 2310, and a 1×1 convolution 2312. In some examples, Figure 20 The parameters include C = 24, C1 = 72, and C 21 = 24.

[0189] Figure 24 This is a block diagram illustrating an example in-loop filter network architecture with luminance / chrominance segmentation according to one or more aspects of this disclosure. For example, Figure 24 A neural network architecture with luminance / chrominance segmentation for 17k MAC / pixel complexity is shown.

[0190] Figure 24 The architecture includes a header block 2480, a fusion block 2422, a set of one or more chroma backbone blocks 2428, a set of one or more luma backbone blocks 2429, a tail block 2486, and a tail block 2488. The inputs to the header block 2480 include IPB2400 and QP... SLICE 2402, QP BASE 2404, BS 2406, PRED 2408 and REC 2409.

[0191] The header block 2480 includes 1×1 convolutions 2410A to 2410D and 3×3 convolutions 2410E to 2410F applied to the corresponding input.

[0192] The fusion block 2422 can use 1×1 convolution 2418 and PReLU 2420 to fuse feature maps. Figure 24 The filter can process the output of PReLU 2420 by applying 3×3 convolution 2424 and PReLU 2426. The set 2428 of chroma backbone blocks 2430A to 2430N can provide chroma samples. Figure 24 The filter function of the filter. The set 2429 of luminance backbone blocks 2432A to 2432N can provide luminance. Figure 24 The filtering function of the filter.

[0193] Tail block 2486 may include the set 2428 located after the chroma backbone blocks 2430A to 2430N. Figure 24 The filter block, and the tail block 2488 may include the set 2429 located after the luminance backbone blocks 2432A to 2432N. Figure 24 The filter block 2486 may include a 1×3 separable convolution 2470, a 3×1 separable convolution 2472, a 1×1 convolution 2450, a PReLU 2452, and a 3×3 convolution 2454. The output of the 3×3 convolution 2454 can be combined with the reconstructed chroma samples and truncated by a clipping unit 2458, which outputs the filtered reconstructed chroma samples.

[0194] The tail block 2488 may include a 1×3 separable convolution 2474, a 3×1 separable convolution 2476, a 1×1 convolution 2460, a PReLU 2462, a 3×3 convolution 2464, and a pixel reconstruction 2456. The output of the pixel reconstruction 2456 can be combined with the reconstructed luminance sample and cropped by a cropping unit 2468, which outputs a filtered reconstructed luminance sample.

[0195] Figure 24 Examples of sample architectures and parameters are shown, such as the number of inputs, features, channels, and the number of residual blocks. In some examples, Figure 24 and / or Figure 25 The example parameters include one or more of the following: d1 = 12, d2 = 8, d3 = 4, d4 = 2, d5 = 2, d6 = 24, N (brightness) = 14, C (brightness) = 16, C1 (brightness) = 64, C 21 (Luminosity) = 16, N (Chromaticity) = 4, C (Chromaticity) = 16, C1 (Chromaticity) = 32 and C 21 (Chroma) = 16. For the illustrated configuration, the complexity is approximately 17kMAC / pixel.

[0196] Figure 25This is an example backbone block according to one or more aspects of this disclosure. Backbone block 2500 may be... Figure 24 Examples of any one of the chroma backbone blocks 2430A to 2430N and / or luma backbone blocks 2432A to 2432N. Backbone block 2500 includes a 1×1 convolution 2502, a PReLU 2504, a 1×1 convolution 2506, a 1×3 separable convolution 2508, a 3×1 separable convolution 2510, and a 1×1 convolution 2512. Backbone block 2500 can combine the output of the 1×1 convolution 2512 with the input of backbone block 2500 to generate the output of backbone block 2500. The above refers to... Figure 24 The example parameters described can be used Figure 25 The main block is 2500.

[0197] To further reduce the complexity to <10kMAC / pixel while retaining most of the decoding performance, an alternative implementation of the UF architecture utilizing regular clustering (CP) decomposition of separable convolutions is described.

[0198] In one example, the multidimensional 3×3 convolution is simplified to a 3-component 1D decomposition and utilized throughout the architecture. Since the 3-stage decomposition employs feature reduction (e.g., C1 < C), it can utilize alternating sequential decomposition via a pair of backbone blocks. For example, the first backbone block could employ a 1×3 convolution with C1 outputs applied first, followed by a 3×1 convolution with feature expansion back to C. The second backbone in this pair could apply a reverse horizontal / vertical decomposition. Activation layers (e.g., PReLU) could be applied in the final stage of the decomposition. Figure 26 A visualization of the proposed low-complexity backbone is shown.

[0199] Figure 26 This is a block diagram illustrating an example low-complexity backbone block according to one or more aspects of this disclosure. Backbone block 2600 may include a 1×1 convolution 2602, a 1×3 separable convolution 2604, a 3×1 separable convolution 2606, and a PReLU 2608. Backbone block 2600 may combine the output of PReLU 2608 with the input of backbone block 2600.

[0200] The backbone block 2600 can input the combination of the output of PReLU 2608 and the input of the backbone block 2600 into a 1×1 convolution 2612, a 3×1 separable convolution 2614, a 1×3 separable convolution 2616, and a PReLU 2618. The backbone block 2600 can combine the output of PReLU 2618 with the combination of the output of PReLU 2608 and the input of the backbone block 2600 to generate the output of the backbone block 2600.

[0201] Figure 27 An alternative design for the main block is presented. Following... Figure 26 Following the same design principle as the example, the activation layer (e.g., PReLU) is placed after the first depth 1×1 convolution.

[0202] Figure 27 This is a block diagram illustrating another example of a low-complexity backbone block according to one or more aspects of this disclosure. Backbone block 2700 may include a 1×1 convolution 2702, a PReLU 2704, a 1×3 separable convolution 2706, and a 3×1 separable convolution 2708. Backbone block 2700 may combine the output of the 3×1 separable convolution 2708 with the input of backbone block 2700.

[0203] The backbone block 2700 can input the combination of the output of the 3×1 separable convolution 2708 and the input of the backbone block 2700 into the 1×1 convolution 2712, PReLU 2714, 3×1 separable convolution 2716, and 1×3 separable convolution 2718. The backbone block 2700 can combine the output of the 1×3 separable convolution 2718 with the combination of the output of the 3×1 separable convolution 2708 and the input of the backbone block 2700 to generate the output of the backbone block 2700.

[0204] Figure 28 This is a block diagram illustrating a specific implementation of an example fusion block according to one or more aspects of this disclosure. The complexity of the fusion / transformation block can be reduced by employing separable convolutions. The downsampling process (denoted as d2) can be implemented at each stage of spatial processing, such as... Figure 28 As shown.

[0205] The fusion block 2800 may include a 1×1 convolution 2802, a PReLU 2804, a 1×3 separable convolution 2806, and a 3×1 separable convolution 2808.

[0206] In addition to employing spatial downsampling at each stage of decomposition, the design of fusion block 2800 can also follow the same design principles as the backbone blocks (e.g., backbone block 2600 or backbone block 2700). In some examples, fusion blocks with different decomposition orders can be used, similar to... Figure 26 and Figure 27 The example shown is a trunk design.

[0207] although Figure 28 The example shows a downsampling (dX) parameter value of 2 for the amount of downsampling in the specified blending block; however, in some examples, the downsampling parameter may have an integer value not equal to 2. In some examples, this downsampling parameter may have different values ​​for the luma component or the chroma component. Alternatively, the blending block can be performed separately for the luma and chroma branches, thus employing the optimal parameter selection for each component (e.g., luma and chroma components). For example, the parameter value for the luma component may not be equal to the parameter value for the chroma component.

[0208] Additionally or alternatively, the complexity of the header block and the number of trunk blocks can be optimized to achieve the target or desired complexity. In some examples, the following parameters can be utilized when the target complexity level is below 5k MAC / pixel. (Reference) Figure 19 The block diagram can use any of the following example configurations:

[0209] Configuration 1: "D1": 8, "D2": 2, "D3": 1, "D4": 1, "D5": 1, "D6": 8, "N_Y":10, "N_UV": 4, "C": 4, "C1_Y": 8, "C1_UV": 8, "C21": 4, which has a wider range of filter support at the cost of a reduced number of channels C=4.

[0210] Configuration 2: "D1": 8, "D2": 2, "D3": 1, "D4": 1, "D5": 1, "D6": 8, "N_Y": 4,"N_UV": 1, "C": 8, "C1_Y": 8, "C1_UV": 8, "C21": 4, which has a higher number of channels at the cost of a reduced number of backbone blocks, with N_Y=4 and N_UV=1.

[0211] To improve the complexity-performance tradeoff in the very low complexity range (e.g., 1kMAC / pixel to 10kMAC / pixel), several aspects of the in-loop filter architecture are described. The video encoder 200 or the video decoder 300 may implement any or any combination of these architectural aspects.

[0212] Figure 29 This is a block diagram illustrating an example restricted header block design according to one or more aspects of this disclosure. In one example, a restricted header block design (e.g., performing input feature extraction) is described. The inputs to header block 2980 include IPB2900, QP... SLICE 2902, QP BASE 2904, BS 2906, PRED 2908, and REC 2909. Header block 2980 includes 1×1 convolutions 2910A to 2910D and 3×3 convolutions 2910E to 2910F applied to the corresponding inputs, as shown in the figure.

[0213] In some examples, in such designs, the sum of the number of output features (e.g., output channels) extracted independently from each input channel is less than or equal to the number of output channels produced by fusing the input features into a single output. Increasing the number of channels introduced into the fused data representation (e.g., expansion) allows for capturing various combinations of context, rather than features per channel.

[0214] For example, the restricted header block 2980 can be configured for each input channel (e.g., IPB 2900, QP). SLICE 2902, QP BASE (2904, BS 2906, PRED 2908, and REC 2909) extract the corresponding number of output channels. For example, a 1×1 convolution 2910A can extract d5 output channels for IPB 2900. A 1×1 convolution 2910B can extract d5 output channels for QP. SLICE 2902 extracts d4 output channels. A 1×1 convolution 2910C can be used for QP. BASE 2904 extracts d4 output channels. A 1×1 convolution 2910D extracts d3 output channels for BS 2906. A 3×3 convolution 2910E extracts d2 output channels for PRED 2908. A 3×3 convolution 2910F extracts d1 output channels for REC 2909. These extractions can be independent, as they can be performed individually for each of the separate inputs. The sum of the corresponding number of output channels can be less than or equal to the number of fused output channels (d6) of a single output 2914.

[0215] For example, the restricted header block 2980 may include multiple input channels and be configured to independently extract a corresponding number of output channels for each of the multiple input channels and fuse the outputs associated with each of the multiple input channels into a single output, wherein the sum of the corresponding number of output channels is less than or equal to the number of fused output channels of the single output.

[0216] In the example architecture, Figure 29 As shown, d6 ≥ sum(d1,d2,d3,d4,d4,d5), where d1 to d5 define the number of output channels independently extracted from the input channels, and d6 defines the number of output channels that have been fused. Example parameters for such a constrained header block design of header block 2980 can be configured as: {d1 = 4, d2 = 2, d3 = 1, d4 = 1, d5 = 1, d6 = 16}, or {d1 = 12, d2 = 8, d3 = 4, d4 = 2, d5 = 2, d6 = 32}. For example, video encoder 200 or video decoder 300 can implement the header block 2980 as an NN-based in-loop filter.

[0217] Figure 30 This is a block diagram illustrating an example sequential application and low-complexity transformation block that incorporates one or more aspects of this disclosure. Figure 30 The modified transform block combined with the fusion block is shown.

[0218] The combined transform block and fusion block 3030 may include a fusion block 3000 and a transform block 3010. The fusion block 3000 may include a 1×1 convolution 3002 and a PReLU 3004. The transform block 3010 may include a 1×3 separable convolution 3012, a 3×1 separable convolution 3014, a 1×1 convolution 3016, and a PReLU 3018.

[0219] In some examples, the decomposed parameters are d6=C1=C2=C3. The presence, placement, and activation type (non-linear, such as ReLU, PReLU, etc.) and / or order of 1×3 separable convolutions and 3×1 separable convolutions can vary depending on the architecture or the properties of the processed signal. The video encoder 200 or video decoder 300 can implement the combined transform block and fusion block 3030 as part of an NN-based in-loop filter.

[0220] Figure 31 This is a block diagram illustrating an example NN-based in-loop filter architecture with luminance / chrominance segmentation according to one or more aspects of this disclosure. Figure 24 The luminance / chrominance segmentation utilized in the example NN architecture shown employs transform blocks with an equal number of outputs representing luminance and chrominance features, such that C Y = C UV To reduce "bottlenecks" (e.g., constraints) in brightness representation, refer to Figure 31 This describes a transformation block with non-equal luminance / chrominance output feature segmentation. For example, making the number of luminance output features different from (e.g., greater than) the number of chrominance output features reduces the constraints on the luminance representation and thus can produce higher quality results than when the number of luminance output features is the same as the number of chrominance output features, such as... Figure 24 As shown in the architecture.

[0221] Figure 31 Examples include transform block 3110, and chroma backbone blocks 3130A to 3130N. C and brightness backbone 3132A to 3132N Y Transformation block 3110 may include 1×3 separable convolutions 3112, 3×1 separable convolutions 3114, 1×1 convolutions 3116, and PReLU 3118. In some examples, the number of luminance features produced as the output of the transformation block (e.g., transformation block 3110) is greater than the number of chrominance (U, V components) features. In some examples, the corresponding parameters may be equal to: {C Y = 24, C UV = 8},{C Y =16, C UV =8},{C Y =12, C UV=4}, or others. For example, a video encoder 200 or a video decoder 300 can implement transform blocks with unequal numbers of luminance output features and chrominance output features. In some examples, the number N of chrominance backbone blocks is... C The number N of the brightness backbone blocks Y Different. In some examples, the number N of chroma backbone blocks... C This can be equal to the number of luminance backbone blocks N. Y .

[0222] For example, a neural network-based filter may include a transformation block having a first number of luma output features and a second number of chroma output features, wherein the first number is greater than the second number. The neural network-based filter may include multiple luma backbone blocks configured to process the luma output features. The neural network-based filter may also include multiple chroma backbone blocks configured to process the chroma output features. In some examples, the first number is equal to 24 and the second number is equal to 8, the first number is equal to 16 and the second number is equal to 8, or the first number is equal to 12 and the second number is equal to 4.

[0223] The techniques disclosed herein can be applied to in-loop filters with a wide range of complexities, for example, by selecting different parameter configurations.

[0224] The techniques disclosed herein reduce computational complexity and memory bandwidth requirements and provide competitive performance for neural network video decoding (NNVC) architectures. The examples described in this document relate to NN-assisted loop filtering; however, such examples are applicable to any NN-based video decoding tool that consumes input data with certain statistical properties, such as static content or sparse representation.

[0225] Figure 32 This is a flowchart illustrating an example of an in-loop filter technique based on an neural network according to one or more aspects of this disclosure. The video decoder 300 may receive video data (3200) indicating a picture. For example, the video decoder 300 may receive video data as a bitstream from the video encoder 200.

[0226] The video decoder 300 can reconstruct blocks of video data to generate reconstructed blocks (3202). For example, the video decoder 300 can decode blocks of video data using decoding modes such as inter-frame prediction or intra-frame prediction.

[0227] The video decoder 300 can perform a neural network-based filtering process on the reconstructed block to generate a filtered block (3204). For example, the video decoder 300 can perform the neural network-based filtering process via a loop filter unit 148. The neural network-based filtering process may include in-loop filtering. The neural network-based filtering process may include applying a neural network-based filter. The neural network-based filter may include a restricted header block having multiple input channels and can be configured to independently extract a corresponding number of output channels for each of the multiple input channels and fuse the outputs associated with each of the multiple input channels into a single output, wherein the sum of the individual numbers of the output channels is less than or equal to the number of fused output channels of the single output.

[0228] For example, the restricted header block 2980 can be configured to independently extract the d5 number of output channels from the IPB 2900, and from the QP SLICE 2902's d4 number of output channels, from QP BASE The output channels 2904 have d4 output channels, BS2906 has d3 output channels, PRED 2908 has d2 output channels, and REC 2909 has d1 output channels. d1 through d5 can therefore each represent the number of output channels. The restricted header block 2980 can also be configured to merge the outputs associated with each of the multiple input channels into a single output 2914. For example, the sum of the corresponding number of output channels (sum(d1, d2, d3, d4, d5, ...)). d5) can be less than or equal to the number of fused output channels of a single output 2914 (d6).

[0229] In some examples, the neural network-based filter further includes a transition block having a first number of luminance output features and a second number of chrominance output features, wherein the first number is greater than the second number. In some examples, the neural network-based filter also includes multiple luminance backbone blocks configured to process the luminance output features and multiple chrominance backbone blocks configured to process the chrominance output features. In some examples, the first number is equal to 24 and the second number is equal to 8, the first number is equal to 16 and the second number is equal to 8, or the first number is equal to 12 and the second number is equal to 4.

[0230] In some examples, the NN-based filter also includes a transformation block (e.g., transformation block 3010) that comprises one or more convolutions and an activation block (e.g., PReLU 3018), the activation block following all the convolutions in the one or more convolutions of the transformation block. In other words, each of the one or more convolutions is applied to the data flowing through the transformation block before the activation block is applied. For example, the activation block may generate the output of the transformation block.

[0231] In some examples, d6 ≥ sum (d1, d2, d3, d4, d5), where d1 to d5 are the number of corresponding output channels (e.g., counts), and d6 is the number of fused output channels for a single output. In some examples, multiple input channels include indications for reconstructed samples, predicted samples, boundary intensity maps, second quantization parameters, and first quantization parameters, and d1 = 4, d2 = 2, d3 = 1, d4 = 1, d5 = 1, and d6 = 16. In some examples, multiple input channels include indications for reconstructed samples, predicted samples, boundary intensity maps, second quantization parameters, and first quantization parameters, and d1 = 12, d2 = 8, d3 = 4, d4 = 2, d5 = 2, and d6 = 32.

[0232] In some examples, the NN-based filter also includes a transformation block combined with the fusion block (e.g., a combined transformation block and fusion block 3030), wherein the fusion block includes a 1×1 convolution block with a first number of channels (d6), wherein the transformation block includes a first separable convolution with a second number of channels (C1), a second separable convolution with a third number of channels (C2), and a third convolution with a fourth number of channels (C3), and wherein d6 = C1 = C2 = C3.

[0233] Figure 33 This is a block diagram illustrating an example video encoder 200 that can perform the techniques of this disclosure. Figure 33 This disclosure is provided for illustrative purposes and should not be construed as a limitation on the techniques extensively illustrated and described herein. For illustrative purposes, this disclosure describes the video encoder 200 in accordance with the techniques of VVC and HEVC. However, the techniques of this disclosure can be performed by video encoding devices configured for other video decoding standards and video decoding formats, such as AV1 and subsequent formats of AV1 video decoding.

[0234] exist Figure 33In the example, 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, mode selection unit 202, residual generation unit 204, transform processing unit 206, quantization unit 208, inverse quantization unit 210, inverse transform processing unit 212, reconstruction unit 214, filter unit 216, DPB 218, and entropy coding unit 220 may be implemented in one or more processors or in processing circuitry. For example, the units of the video encoder 200 may be implemented as one or more circuit or logic elements as part of hardware circuitry, or as part of a processor, ASIC, or FPGA. Furthermore, the video encoder 200 may include additional or alternative processors or processing circuitry to perform these and other functions.

[0235] Video data storage 230 can store video data to be encoded by components of video encoder 200. Video encoder 200 can receive data from, for example, video source 104 (…). Figure 1 The video data memory 230 receives video data stored in the video data memory 230. The DPB 218 can act as a reference picture memory, storing reference video data for use when the video encoder 200 predicts subsequent video data. The video data memory 230 and DPB 218 can be formed from 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 DPB 218 can be provided by the same memory device or separate memory devices. In various examples, the video data memory 230 can be on-chip (as illustrated) with other components of the video encoder 200, or off-chip relative to those components.

[0236] In this disclosure, references to video data memory 230 should not be construed as limited to memory inside video encoder 200 (unless specifically described) or memory outside video encoder 200 (unless specifically described). Rather, references to video data memory 230 should be understood as a reference memory that stores video data received by video encoder 200 for encoding (e.g., video data for the current block to be encoded). Figure 1 The memory 106 can also provide temporary storage for the outputs from various units of the video encoder 200.

[0237] Examples Figure 33Various units help understand the operations performed by the video encoder 200. Units can be implemented as fixed-function circuits, programmable circuits, or combinations thereof. Fixed-function circuits are circuits that provide specific functionality and are pre-defined for the operations that can be performed. Programmable circuits are circuits that can be programmed to perform various tasks and provide flexible functionality for 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 or output parameters), but the type of operation performed by a fixed-function circuit is generally immutable. In some examples, one or more units in the unit may be different circuit blocks (fixed-function or programmable), and in some examples, one or more units in the unit may be integrated circuits.

[0238] The video encoder 200 may include an arithmetic logic unit (ALU), an essential function unit (EFU), digital circuitry, analog circuitry, and / or a programmable core, all formed by programmable circuitry. In an example where the operation of the video encoder 200 is performed using software executed by programmable circuitry, memory 106 ( Figure 1 The video encoder 200 may store instructions (e.g., target code) of the software received and executed by the video encoder 200, or another memory (not shown) within the video encoder 200 may store such instructions.

[0239] The video data storage unit 230 is configured to store received video data. The video encoder 200 can retrieve images of the video data from the video data storage unit 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 storage unit 230 can be raw video data to be encoded.

[0240] The mode selection unit 202 includes a motion estimation unit 222, a motion compensation unit 224, and an intra-frame prediction unit 226. The mode selection unit 202 may include additional functional units for performing video prediction based on other prediction modes. As an example, the mode selection unit 202 may include a palette unit, an intra-frame block copying 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.

[0241] Mode selection unit 202 typically coordinates multiple coding channels to test combinations of coding parameters and the resulting rate-distortion values ​​for such combinations. Coding parameters may include the CTU-CU partitioning, the prediction mode for the CU, the transformation type of the residual data for the CU, the quantization parameters of the residual data for the CU, etc. Mode selection unit 202 can ultimately select a combination of coding parameters that has a better rate-distortion value compared to other tested combinations.

[0242] The video encoder 200 can divide an image retrieved from the video data storage 230 into a series of CTUs and encapsulate one or more CTUs within a slice. The mode selection unit 202 can divide the image's CTUs according to the tree structure described above (such as an MTT structure, QTBT structure, superblock structure, or the quadtree structure described above). As described above, the video encoder 200 can 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".

[0243] Generally, mode selection unit 202 also controls its components (e.g., motion estimation unit 222, motion compensation unit 224, and intra-prediction unit 226) to generate prediction blocks for the current block (e.g., the current CU, or, in HEVC, the overlapping portion of PU and TU). To perform inter-frame prediction for 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 decoded pictures stored in DPB 218). Specifically, motion estimation unit 222 may calculate values ​​representing the similarity between a potential reference block and the current block, for example, based on sum of absolute differences (SAD), sum of squared differences (SSD), mean absolute difference (MAD), mean squared difference (MSD), etc. Motion estimation unit 222 may typically perform these calculations using sample-by-sample differences between the current block and the reference blocks under consideration. Motion estimation unit 222 may identify reference blocks with the lowest values ​​produced by these calculations to indicate the reference block that best matches the current block.

[0244] Motion estimation unit 222 can generate one or more motion vectors (MVs) that define the location of a reference block in a reference image relative to the location of the current block in the current image. Motion estimation unit 222 can then provide the motion vectors to motion compensation unit 224. For example, for unidirectional inter-frame prediction, motion estimation unit 222 can provide a single motion vector, while for bidirectional inter-frame prediction, motion estimation unit 222 can provide two motion vectors. Motion compensation unit 224 can then use the motion vectors to generate a prediction block. For example, motion compensation unit 224 can use the motion vectors to retrieve data for the reference block. As another example, if the motion vectors have fractional sample accuracy, motion compensation unit 224 can interpolate the prediction block according to one or more interpolation filters. Furthermore, for bidirectional inter-frame prediction, motion compensation unit 224 can retrieve data for two reference blocks identified by corresponding motion vectors and combine the retrieved data, for example, by per-sample averaging or weighted averaging.

[0245] When operating according to the AV1 video decoding format, the motion estimation unit 222 and the motion compensation unit 224 can be configured to encode the decoded blocks of video data (e.g., both luma and chroma decoded blocks) using translational motion compensation, affine motion compensation, overlap block motion compensation (OBMC), and / or composite inter-intra-frame prediction.

[0246] As another example, for intra-prediction or intra-prediction decoding, intra-prediction unit 226 may generate a prediction block from samples adjacent to the current block. For example, for directional mode, intra-prediction unit 226 may typically mathematically combine the values ​​of adjacent samples and fill these calculated values ​​across the current block in a defined direction to produce a prediction block. As another example, for DC mode, intra-prediction unit 226 may calculate the average of the adjacent samples of the current block and generate a prediction block to include the resulting average for each sample of the prediction block.

[0247] When operating according to the AV1 video decoding format, the intra-frame prediction unit 226 can be configured to encode decoded blocks of video data (e.g., both luma and chroma decoded blocks) using directional intra-frame prediction, non-directional intra-frame prediction, recursive filter intra-frame prediction, luma-chroma (CFL) prediction, intra-block copying (IBC), and / or palette modes. The mode selection unit 202 may include additional functional units for performing video prediction based on other prediction modes.

[0248] Mode selection unit 202 provides a prediction block to residual generation unit 204. Residual generation unit 204 receives an uncoded raw version of the current block from video data memory 230 and a 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 the residual block for the current block. In some examples, residual generation unit 204 may also determine the differences between sample values ​​in the residual block to generate the residual block using residual differential pulse decoding modulation (RDPCM). In some examples, residual generation unit 204 may be formed using one or more subtractor circuits performing binary subtraction.

[0249] In the example where mode selection unit 202 divides a CU into PUs, each PU can be associated with a luma prediction unit and a corresponding chroma prediction unit. Video encoder 200 and video decoder 300 can support PUs of various sizes. As noted above, the size of a CU can refer to the size of its luma decoding block, while the size of a PU can refer to the size of its luma prediction unit. Assuming a particular CU size is 2N×2N, video encoder 200 can support PU sizes of 2N×2N or N×N for intra-frame prediction, and symmetric PU sizes of 2N×2N, 2N×N, N×2N, N×N, or similar for inter-frame prediction. Video encoder 200 and video decoder 300 can also support asymmetric partitioning for PU sizes of 2N×nU, 2N×nD, nL×2N, and nR×2N for inter-frame prediction.

[0250] In an example where mode selection unit 202 does not further divide the CU into PUs, each CU can be associated with a luminance decoding block and a corresponding chrominance decoding block. As mentioned above, the size of the CU can refer to the size of the luminance decoding block of the CU. The video encoder 200 and the video decoder 300 can support CU sizes of 2N×2N, 2N×N, or N×2N.

[0251] For other video decoding techniques, such as intra-block copy mode decoding, affine mode decoding, and linear model (LM) mode decoding, as some examples, mode selection unit 202 generates a prediction block for the current block being encoded via a corresponding unit associated with the decoding technique. In some examples (such as palette mode decoding), mode selection unit 202 may not generate a prediction block, but instead generate syntax elements indicating how the block is reconstructed based on a selected palette. In such modes, mode selection unit 202 may provide these syntax elements to entropy coding unit 220 for encoding.

[0252] As described above, the residual generation unit 204 receives video data for the current block and the corresponding prediction block. Then, the residual generation unit 204 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.

[0253] Transform processing unit 206 applies one or more transformations to the residual block to generate a block of transform coefficients (referred to herein as a "transform coefficient block"). Transform processing unit 206 may apply various transformations to the residual block to form the transform coefficient block. For example, transform processing unit 206 may apply a discrete cosine transform (DCT), direction transform, Karhunen-Loeve transform (KLT), or conceptually similar transformations to the residual block. In some examples, transform processing unit 206 may perform multiple transformations on the residual block, such as primary and secondary transformations (e.g., rotation transformations). In some examples, transform processing unit 206 does not apply any transformations to the residual block.

[0254] When operating according to AV1, 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"). Transform processing unit 206 may apply various transforms to the residual block to form the transform coefficient block. For example, transform processing unit 206 may apply a combination of horizontal / vertical transforms, which may include the Discrete Cosine Transform (DCT), the Asymmetric Discrete Sine Transform (ADST), the Reversed ADST (e.g., ADST in reverse order), and the Identity Transform (IDTX). When using the Identity Transform, the transform is skipped in either the vertical or horizontal direction. In some examples, transform processing may be skipped entirely.

[0255] Quantization unit 208 quantizes the transform coefficients in a transform coefficient block to produce a quantized transform coefficient block. Quantization unit 208 quantizes the transform coefficients of the transform coefficient block according to the quantization parameter (QP) value associated with the current block. Video encoder 200 (e.g., via mode selection unit 202) can 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 cause information loss, and therefore, the quantized transform coefficients may have lower accuracy compared to the original transform coefficients produced by transform processing unit 206.

[0256] The inverse quantization unit 210 and the inverse transform processing unit 212 can apply inverse quantization and inverse transform, respectively, to the quantized transform coefficient block to reconstruct the residual block based on the transform coefficient block. The reconstruction unit 214 can generate a reconstructed block corresponding to the current block (although potentially 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 can add samples of the reconstructed residual block to corresponding samples of the prediction block generated by the mode selection unit 202 to generate the reconstructed block.

[0257] Filter unit 216 may perform one or more filtering operations on the reconstructed block. For example, filter unit 216 may perform a deblocking operation to reduce block 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 decoding techniques described above. For example, filter unit 216 may be configured to receive reconstructed blocks of images of video data and perform a NN-based filtering process on the reconstructed blocks to generate filtered blocks. The NN-based filtering process includes performing multiple separable convolutions to approximate multidimensional convolutions.

[0258] When operating according to AV1, filter unit 216 may perform one or more filtering operations on the reconstructed block. For example, filter unit 216 may perform a deblocking operation to reduce block artifacts along the edges of the CU. In other examples, filter unit 216 may apply a constrained direction enhancement filter (CDEF) after deblocking and may include the application of a non-separable, nonlinear, low-pass directional filter based on the estimated edge direction. Filter unit 216 may also include a loop recovery filter applied after CDEF and may include a separable symmetric normalized Wiener filter or a dual-guided filter.

[0259] The video encoder 200 stores reconstructed blocks in the DPB 218. For example, in an example where the filter unit 216 is not operated, the reconstruction unit 214 may store reconstructed blocks in the DPB 218. In an example where the filter unit 216 is operated, the filter unit 216 may store filtered reconstructed blocks in the DPB 218. The motion estimation unit 222 and the motion compensation unit 224 may retrieve reference images formed by the reconstructed (and potentially filtered) blocks from the DPB 218 to perform inter-frame prediction for blocks of subsequent encoded images. Additionally, the intra-frame prediction unit 226 may use the reconstructed blocks of the current image in the DPB 218 to perform intra-frame prediction for other blocks in the current image.

[0260] Generally, entropy coding unit 220 can entropy code syntax elements received from other functional components of video encoder 200. For example, entropy coding unit 220 can entropy code quantized transform coefficient blocks from quantization unit 208. As another example, entropy coding unit 220 can entropy code predictive 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 can perform one or more entropy coding operations on syntax elements (another example of video data) to generate entropy-coded data. For example, entropy coding unit 220 can perform context-adaptive variable-length decoding (CAVLC), CABAC, variable-to-variable (V2V) length decoding, syntax-based context-adaptive binary arithmetic decoding (SBAC), probability interval partitioning entropy (PIPE) decoding, exponential Golomb coding, or another type of entropy coding operation on the data. In some examples, entropy coding unit 220 can operate in a bypass mode where syntax elements are not entropy-coded.

[0261] The video encoder 200 can output a bitstream that includes the entropy coding syntax elements required to reconstruct slices or blocks of images. Specifically, the entropy coding unit 220 can output a bitstream.

[0262] According to AV1, entropy coding unit 220 can be configured as a symbol-to-symbol adaptive multi-symbol arithmetic decoder. The syntax elements in AV1 consist of an N-element alphabet, and the context (e.g., a probability model) consists of a set of N probabilities. Entropy coding unit 220 can store the probabilities as an n-bit (e.g., 15-bit) cumulative distribution function (CDF). Entropy coding unit 220 can perform recursive scaling using an update factor based on the alphabet size to update the context.

[0263] The operations described above are relative to blocks. This description should be understood as operations applied to luma decoding blocks and / or chroma decoding blocks. As described above, in some examples, the luma decoding block and chroma decoding block are the luma and chroma components of the CU. In some examples, the luma decoding block and chroma decoding block are the luma and chroma components of the PU.

[0264] In some examples, it is not necessary to repeat the operations performed relative to the luma decoder for the chroma decoder block. As an example, the operations for identifying the motion vector (MV) and reference image of the luma decoder block do not need to repeat the MV and reference image used to identify the chroma block. Instead, the MV used for the luma decoder block can be scaled to determine the MV used for the chroma block, and the reference image can be the same. As another example, the intra-frame prediction process can be the same for both the luma and chroma decoders.

[0265] Video encoder 200 represents an example of a device configured to encode video data, the device including: one or more memories configured to store images of video data; and one or more processors communicating with the one or more memories, the one or more processors being configured to: determine video data indicating the images; reconstruct blocks of the video data to generate reconstructed blocks; and perform a neural network (NN)-based filtering process on the reconstructed blocks to generate filtered blocks, wherein the NN-based filtering process includes applying an NN-based filter, the NN-based filter including a restricted head block having multiple input channels and configured to independently extract a corresponding number of output channels for each of the multiple input channels and fuse the outputs associated with each of the multiple input channels into a single output, wherein the sum of the corresponding number of output channels is less than or equal to the number of fused output channels of the single output.

[0266] Figure 34 This is a block diagram illustrating an example video decoder 300 that can perform the techniques of this disclosure. Figure 34 This disclosure is provided for illustrative purposes and not for limiting the techniques extensively illustrated and described herein. For illustrative purposes, the video decoder 300 is described in accordance with VVC and HEVC techniques. However, the techniques of this disclosure can be implemented by video decoding devices configured for other video decoding standards.

[0267] exist Figure 34 In the example, the video decoder 300 includes a decoded 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, entropy decoding unit 302, prediction processing unit 304, inverse quantization unit 306, inverse transform processing unit 308, reconstruction unit 310, filter unit 312, and DPB 314 can be implemented in one or more processors or in processing circuitry. For example, the units of the video decoder 300 can be implemented as one or more circuit or logic elements as part of hardware circuitry, or as part of a processor, ASIC, or FPGA. Furthermore, the video decoder 300 may include additional or alternative processors or processing circuitry to perform these and other functions.

[0268] 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 that perform predictions based on other prediction modes. As an example, the prediction processing unit 304 may include a palette unit, an intra-block copying 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.

[0269] When operating according to AV1, motion compensation unit 316 can be configured to decode video data blocks (e.g., both luma and chroma blocks) using translational motion compensation, affine motion compensation, OBMC, and / or composite inter-intra-frame prediction, as described above. Intra-frame prediction unit 318 can be configured to decode video data blocks (e.g., both luma and chroma 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.

[0270] CPB memory 320 can store video data to be decoded by components of video decoder 300, such as encoded video bitstreams. For example, it can be retrieved from computer-readable medium 110 ( Figure 1 The video data stored in the CPB memory 320 is obtained. The CPB memory 320 may include a CPB that stores encoded video data (e.g., syntax elements) from the encoded video bitstream. Furthermore, the CPB memory 320 may store video data other than the syntax elements of the decoded picture, such as temporary data representing the output from various units of the video decoder 300. The DPB 314 typically stores a decoded picture that the video decoder 300 may output, and / or uses the decoded picture as reference video data when decoding subsequent data or pictures of the encoded video bitstream. The CPB memory 320 and the DPB 314 may be formed of any memory device from various 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 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.

[0271] Additionally or alternatively, in some examples, the video decoder 300 may be from the memory 120 ( Figure 1The decoded video data can be retrieved from the memory. In other words, memory 120 can utilize CPB memory 320 to store data as discussed above. Similarly, when some or all of the functionality of video decoder 300 is implemented in software to be executed by the processing circuitry of video decoder 300, memory 120 can store instructions to be executed by video decoder 300.

[0272] Examples Figure 34 The various units shown help to understand the operations performed by the video decoder 300. These units can be implemented as fixed-function circuits, programmable circuits, or a combination thereof. Similar to... Figure 33 Fixed-function circuits are circuits that provide specific functionality and are pre-defined for the operations they can perform. Programmable circuits are circuits that can be programmed to perform various tasks and provide flexible functionality for the operations they can perform. 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 or output parameters), but the type of operation performed by a fixed-function circuit is typically immutable. In some examples, one or more units in a cell may be different circuit blocks (fixed-function or programmable), and in some examples, one or more units in a cell may be integrated circuits.

[0273] The video decoder 300 may include an ALU, EFU, digital circuitry, analog circuitry, and / or a programmable core formed by programmable circuitry. In an example where the operation of the video decoder 300 is performed by software executed on programmable circuitry, on-chip or off-chip memory may store instructions (e.g., object code) of the software received and executed by the video decoder 300.

[0274] Entropy decoding unit 302 can receive encoded video data from the CPB and perform entropy decoding on the video data to reproduce syntax elements. Prediction processing unit 304, inverse quantization unit 306, inverse transform processing unit 308, reconstruction unit 310, and filter unit 312 can generate decoded video data based on syntax elements extracted from the bitstream. Filter unit 312 can be configured to perform any of the aforementioned NN-based video decoding techniques.

[0275] Generally speaking, the video decoder 300 reconstructs the image block by block. The video decoder 300 can perform the reconstruction operation on each block individually (where the block currently being reconstructed (i.e., decoded) can be referred to as the "current block").

[0276] Entropy decoding unit 302 can entropy decode the syntax elements of the quantized transform coefficients defining the quantized transform coefficient block, as well as transform information (such as quantization parameters (QP) and / or transform mode indications). Inverse quantization unit 306 can use the QP associated with the quantized transform coefficient block to determine the degree of quantization, and similarly, determine the degree of inverse quantization to be applied by inverse quantization unit 306. Inverse quantization unit 306 can, for example, perform a bit-by-bit left shift operation to inverse quantize the quantized transform coefficients. Inverse quantization unit 306 can thereby form a transform coefficient block including the transform coefficients.

[0277] After the inverse quantization unit 306 forms the transform coefficient block, the 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, the inverse transform processing unit 308 may apply an inverse DCT, an inverse integer transform, an inverse Karhunen-Loeve transform (KLT), an inverse rotation transform, an inverse direction transform, or another inverse transform to the transform coefficient block.

[0278] Furthermore, the prediction processing unit 304 generates a prediction block based on the prediction information syntax elements entropy decoded by the entropy decoding unit 302. For example, if the prediction information syntax elements indicate that the current block is an inter-frame prediction, the motion compensation unit 316 can generate the prediction block. In this case, the prediction information syntax elements may indicate the reference picture from which the reference block is to be retrieved in the DPB 314, and a motion vector identifying the position of the reference block in the reference picture relative to the position of the current block in the current picture. The motion compensation unit 316 may generally follow the same procedure as relative to the motion compensation unit 224 ( Figure 33 The method described is essentially the same as the method used to perform the inter-frame prediction process.

[0279] As another example, when the prediction information syntax element indicates that the current block is intra-predictive, intra-predictive unit 318 may generate a prediction block according to the intra-predictive mode indicated by the prediction information syntax element. Similarly, intra-predictive unit 318 may generally follow the same procedure as relative to intra-predictive unit 226 ( Figure 33 The intra-prediction process is performed in a manner substantially similar to that described above. The intra-prediction unit 318 can retrieve data of neighboring samples of the current block from the DPB 314.

[0280] Reconstruction unit 310 can use the prediction block and the residual block to reconstruct the current block. For example, reconstruction unit 310 can add samples from the residual block to the corresponding samples from the prediction block to reconstruct the current block.

[0281] Filter unit 312 may perform one or more filtering operations on the reconstructed block. For example, filter unit 312 may perform a deblocking operation to reduce block artifacts along the edges of the reconstructed block. The operation of filter unit 312 is not necessarily performed in all examples. Filter unit 312 may be configured to perform any of the NN-based video decoding techniques described above. For example, filter unit 216 may be configured to receive a reconstructed block of a video image and perform a NN-based filtering process on the reconstructed block to generate a filtered block. The NN-based filtering process includes performing multiple separable convolutions to approximate multidimensional convolutions.

[0282] The video decoder 300 can store reconstructed blocks in the DPB 314. For example, in an example where the operation of the filter unit 312 is not performed, the reconstruction unit 310 can store the reconstructed blocks in the DPB 314. In an example where the operation of the filter unit 312 is performed, the filter unit 312 can store the filtered reconstructed blocks in the DPB 314. As discussed above, the DPB 314 can provide reference information (such as samples of the current image for intra-frame prediction and samples of previously decoded images for subsequent motion compensation) to the prediction processing unit 304. Furthermore, the video decoder 300 can output decoded images (e.g., decoded video) from the DPB 314 for use in applications such as... Figure 1 The subsequent presentation on display devices such as display device 118.

[0283] In this way, video decoder 300 represents an example of a video decoding device, which includes: a memory configured to store video data; and one or more processing units implemented in circuitry and configured to perform NN-based video decoding, including NN-based filtering using any combination of the techniques described above.

[0284] Figure 35 This is a flowchart illustrating an example method for encoding the current block according to the technology of this disclosure. The current block may be or may include the current CU. Although this relates to video encoder 200 ( Figure 1 and Figure 33 This is a description, but it should be understood that other devices can be configured to perform the same actions as described above. Figure 35 Similar to the method.

[0285] In this example, the video encoder 200 initially 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 compute a residual block for the current block (352). To compute the residual block, the video encoder 200 may compute the difference between the unencoded original block for the current block and the prediction 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 use CAVLC or CABAC to encode the transform coefficients. The video encoder 200 may then output the entropy-encoded data of the block (360).

[0286] Figure 36 This is a flowchart illustrating an example method for decoding a current block of video data according to the technology of this disclosure. The current block may be or may include the current CU. Although this relates to video decoder 300 ( Figure 1 and Figure 34 This description is provided, but it should be understood that other devices can be configured to perform similar actions. Figure 36 Similar to the method.

[0287] The video decoder 300 may receive entropy-coded data of the current block, such as entropy-coded prediction information and entropy-coded data for the transform coefficients of the residual block corresponding to the current block (370). The video decoder 300 may entropy decode the entropy-coded 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, for example, using an intra-frame prediction mode or an inter-frame prediction mode indicated by the prediction information of the current block (374), to compute 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 finally decode the current block by combining the prediction block and the residual block (380).

[0288] The following numbered clauses illustrate one or more aspects of the devices and technologies described in this disclosure.

[0289] Aspect 1A. A method for decoding video data, the method comprising: receiving images of video data; reconstructing blocks of the images of the video data to generate reconstructed blocks; and performing a neural network (NN)-based filtering process on the reconstructed blocks to generate filtered blocks, wherein the NN-based filtering process includes applying an NN-based filter, the NN-based filter including a restricted head block having a sum of a number of output channels independently extracted from each input channel, the sum being less than or equal to the number of fused output channels generated by fusing input features into a single output.

[0290] Aspect 2A. The method according to aspect 1A, wherein d6 ≥ sum (d1, d2, d3, d4, d5), where d1 to d5 are the number of output channels extracted independently from each input channel, and d6 is the number of output channels that have been fused.

[0291] Aspect 3A. The method described in aspect 2A, wherein d1 = 4, d2 = 2, d3 = 1, d4 = 1, d5 = 1, and d6 = 16.

[0292] Aspect 4A. The method described in aspect 2A, wherein d1 = 12, d2 = 8, d3 = 4, d4 = 2, d5 = 2, and d6 = 32.

[0293] Aspect 5A. The method according to any one of Aspects 1A to 4A, wherein the NN-based filter further comprises a transformation block combined with a fusion block.

[0294] Aspect 6A. The method described in aspect 5A, wherein the parameters of the decomposition are d6=C1=C2=C3.

[0295] Aspect 7A. The method according to any one of Aspects 1A to 6A, wherein the NN-based filter further comprises luminance and chrominance transition block segmentation, wherein the number of luminance output features from the transition block is greater than the number of chrominance output features from the transition block.

[0296] Aspect 8A. The method according to aspect 7A, wherein the parameters include C Y = 24 and C UV = 8, C Y =16 and C UV =8, or C Y =12 and C UV =4.

[0297] Aspect 9A. An apparatus configured to decode video data, the apparatus comprising: one or more memories configured to store images of the video data; and one or more processors communicating with the one or more memories, the one or more processors being configured to perform the method according to any one of Aspects 1A to 8A.

[0298] Aspect 10A. The apparatus according to aspect 9A, wherein the apparatus is configured to encode video data, and wherein the apparatus further comprises: a camera configured to capture the image of the video data.

[0299] Aspect 11A. The apparatus according to aspect 9A or aspect 10A, wherein the apparatus is configured to decode video data, and wherein the apparatus further comprises: a display configured to display the image of the video data.

[0300] Aspect 12A. An apparatus configured to decode video data, the apparatus comprising at least one component for performing the method according to any one of aspects 1A to 8A.

[0301] Aspect 13A. A non-transitory computer-readable storage medium comprising instructions that, when executed by one or more processors, cause the one or more processors to perform the method according to any one of Aspects 1A to 8A.

[0302] Aspect 1B. A method for decoding video data, the method comprising: receiving video data indicating an image; reconstructing blocks of the video data to generate reconstructed blocks; and performing a neural network (NN)-based filtering process on the reconstructed blocks to generate filtered blocks, wherein the NN-based filtering process includes applying an NN-based filter, the NN-based filter including a restricted header block having a plurality of input channels and configured to independently extract a corresponding number of output channels for each of the plurality of input channels and fuse the outputs associated with each of the plurality of input channels into a single output, wherein the sum of the corresponding number of output channels is less than or equal to the number of fused output channels of the single output.

[0303] Aspect 2B. According to the method of aspect 1B, wherein the NN-based filter further comprises: a transformation block having a first number of luminance output features and a second number of chrominance output features, wherein the first number is greater than the second number; a plurality of luminance backbone blocks configured to process the luminance output features; and a plurality of chrominance backbone blocks configured to process the chrominance output features.

[0304] Aspect 3B. The method according to aspect 2B, wherein the first quantity is equal to 24 and the second quantity is equal to 8, the first quantity is equal to 16 and the second quantity is equal to 8, or the first quantity is equal to 12 and the second quantity is equal to 4.

[0305] Aspect 4B. The method according to aspect 1B, wherein the NN-based filter further includes a transition block and an activation block, the transition block comprising one or more convolutions, the activation block being located after all convolutions in the one or more convolutions of the transition block.

[0306] Aspect 5B. The method according to any one of Aspects 1B to 4B, wherein d6 ≥ sum (d1, d2, d3, d4, d5, ... d5), where d1 to d5 are the number of corresponding output channels, and d6 is the number of fused output channels of the single output.

[0307] Aspect 6B. The method according to aspect 5B, wherein the plurality of input channels include indications of reconstructed samples, predicted samples, boundary intensity maps, second quantization parameters and first quantization parameters, and wherein d1 = 4, d2 = 2, d3 = 1, d4 = 1, d5 = 1, and d6 = 16.

[0308] Aspect 7B. The method according to aspect 5B, wherein the plurality of input channels include indications of reconstructed samples, predicted samples, boundary intensity maps, second quantization parameters and first quantization parameters, and wherein d1 = 12, d2 = 8, d3 = 4, d4 = 2, d5 = 2, and d6 = 32.

[0309] Aspect 8B. According to the method of aspect 1B, the NN-based filter further includes a transformation block combined with a fusion block, wherein the fusion block includes a 1×1 convolution block having a first number of channels (d6), wherein the transformation block includes a first separable convolution having a second number of channels (C1), a second separable convolution having a third number of channels (C2), and a third convolution having a fourth number of channels (C3), and wherein d6 = C1 = C2 = C3.

[0310] Aspect 9B. An apparatus for decoding video data, the apparatus comprising: one or more memories configured to store images of video data; and one or more processors communicating with the one or more memories, the one or more processors configured to: receive video data indicating the images; reconstruct blocks of the video data to generate reconstructed blocks; and perform a neural network (NN)-based filtering process on the reconstructed blocks to generate filtered blocks, wherein the NN-based filtering process includes applying an NN-based filter, the NN-based filter including a restricted head block having a plurality of input channels and configured to independently extract a corresponding number of output channels for each of the plurality of input channels and fuse the outputs associated with each of the plurality of input channels into a single output, wherein the sum of the corresponding number of output channels is less than or equal to the number of fused output channels of the single output.

[0311] Aspect 10B. The device according to aspect 9B, wherein the NN-based filter further comprises: a transition block having a first number of luminance output features and a second number of chrominance output features, wherein the first number is greater than the second number; a plurality of luminance backbone blocks configured to process the luminance output features; and a plurality of chrominance backbone blocks configured to process the chrominance output features.

[0312] Aspect 11B. The device according to aspect 10B, wherein the first quantity is equal to 24 and the second quantity is equal to 8, the first quantity is equal to 16 and the second quantity is equal to 8, or the first quantity is equal to 12 and the second quantity is equal to 4.

[0313] Aspect 12B. The apparatus according to aspect 9B, wherein the NN-based filter further includes a transition block and an activation block, the transition block comprising one or more convolutions, the activation block being located after all the convolutions in the one or more convolutions of the transition block.

[0314] Aspect 13B. The equipment according to any one of Aspects 9B to 12B, wherein d6 ≥ sum (d1, d2, d3, d4, d5, ... d5), where d1 to d5 are the number of corresponding output channels, and d6 is the number of fused output channels of the single output.

[0315] Aspect 14B. The device according to aspect 13B, wherein the plurality of input channels include indications of reconstructed samples, predicted samples, boundary intensity maps, second quantization parameters and first quantization parameters, and wherein d1 = 4, d2 = 2, d3 = 1, d4 = 1, d5 = 1, and d6 = 16.

[0316] Aspect 15B. The device according to aspect 13B, wherein the plurality of input channels include indications of reconstructed samples, predicted samples, boundary intensity maps, second quantization parameters and first quantization parameters, and wherein d1 = 12, d2 = 8, d3 = 4, d4 = 2, d5 = 2, and d6 = 32.

[0317] Aspect 16B. The apparatus according to aspect 9B, wherein the NN-based filter further comprises a transform block combined with a fusion block, wherein the fusion block comprises a 1×1 convolution block having a first number of channels (d6), wherein the transform block comprises a first separable convolution having a second number of channels (C1), a second separable convolution having a third number of channels (C2), and a third convolution having a fourth number of channels (C3), and wherein d6 = C1 = C2 = C3.

[0318] Aspect 17B. The device according to any one of aspects 9B to 16B, wherein the device is configured to decode video data, and wherein the device further includes a display configured to display the image.

[0319] Aspect 18B. An apparatus for encoding video data, the apparatus comprising: one or more memories configured to store images of video data; and one or more processors communicating with the one or more memories, the one or more processors configured to: determine video data indicating the images; reconstruct blocks of the video data to generate reconstructed blocks; and perform a neural network (NN)-based filtering process on the reconstructed blocks to generate filtered blocks, wherein the NN-based filtering process includes applying an NN-based filter, the NN-based filter including a restricted head block having a plurality of input channels and configured to independently extract a corresponding number of output channels for each of the plurality of input channels and fuse the outputs associated with each of the plurality of input channels into a single output, wherein the sum of the corresponding number of output channels is less than or equal to the number of fused output channels of the single output.

[0320] Aspect 19B. The apparatus according to aspect 18B, wherein the NN-based filter further comprises: a transition block having a first number of luminance output features and a second number of chrominance output features, wherein the first number is greater than the second number; a plurality of luminance backbone blocks configured to process the luminance output features; and a plurality of chrominance backbone blocks configured to process the chrominance output features.

[0321] Aspect 20B. The apparatus of claim 18B, wherein the NN-based filter further comprises a transition block and an activation block, the transition block comprising one or more convolutions, the activation block being located after all the convolutions in the one or more convolutions of the transition block.

[0322] It should be recognized that, based on the examples, certain actions or events of any technique described herein may be performed in a different sequence, and may be added, combined, or omitted entirely (e.g., not all actions or events described are necessary for implementing the technique). Furthermore, in some examples, actions or events may be performed concurrently (e.g., through multithreading, interrupt handling, or multiple processors) rather than sequentially.

[0323] In one or more examples, the described functionality may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functionality may be stored as one or more instructions or code on a computer-readable medium or transmitted via a computer-readable medium and executed by a hardware-based processing unit. A computer-readable medium may include a computer-readable storage medium (which corresponds to a tangible medium such as a data storage medium) or a communication medium, including, for example, any medium that facilitates the transfer of a computer program from one place to another according to a communication protocol. In this way, a computer-readable medium 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. A data storage medium may be any available medium that can be accessed by one or more computers or one or more processors to retrieve instructions, code, and / or data structures for implementing the techniques described in this disclosure. Computer program products may include computer-readable media.

[0324] By way of example, and not limitation, such computer-readable storage media may include one or more of RAM, ROM, EEPROM, CD-ROM or other optical disc storage devices, magnetic disk storage devices or other magnetic storage devices, flash memory, or any other medium capable of storing desired program code in the form of instructions or data structures and accessible by a computer. Furthermore, any connection is appropriately referred to as a computer-readable medium. For example, if instructions are transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technology (such as infrared, radio, and microwave), then coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technology (such as infrared, radio, and microwave) is 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 transient media, but instead refer to non-transient tangible storage media. As used herein, disks and optical discs include compact optical discs (CDs), laser optical discs, optical discs, digital versatile optical discs (DVDs), floppy disks, and Blu-ray discs, where disks typically reproduce data magnetically, while optical discs use lasers to reproduce data optically. Combinations of these should also be included within the scope of computer-readable media.

[0325] Instructions can 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. Therefore, the terms "processor" and "processing circuit" as used herein can refer to any of the foregoing structures or any other structure suitable for implementing the techniques described herein. Additionally, in some aspects, the functionality described herein can be provided within dedicated hardware and / or software modules configured for encoding and decoding, or incorporated into combined codecs. Furthermore, these techniques can be fully implemented in one or more circuit or logic elements.

[0326] The techniques disclosed herein can be implemented in a wide variety of devices or apparatuses, including wireless mobile phones, integrated circuits (ICs), or IC sets (e.g., chipsets). Various components, modules, or units are described in this disclosure to emphasize functional aspects of a device configured to perform the disclosed techniques, but implementation by different hardware units is not necessarily required. Rather, as described above, various units can be combined in a codec hardware unit, or various units can be provided by a collection of interoperable hardware units (including one or more processors as described above) combined with appropriate software and / or firmware.

[0327] 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 video data from the indicated image; Reconstruct blocks of the 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 applying an NN-based filter, the NN-based filter including a restricted header block having multiple input channels and configured to independently extract a corresponding number of output channels for each of the multiple input channels and fuse the outputs associated with each of the multiple input channels into a single output, wherein the sum of the corresponding number of output channels is less than or equal to the number of fused output channels of the single output.

2. The method according to claim 1, wherein the NN-based filter further comprises: A conversion block having a first number of luminance output features and a second number of chrominance output features, wherein the first number is greater than the second number; Multiple luminance backbone blocks, the multiple luminance backbone blocks being configured to process the luminance output features; and Multiple chroma backbone blocks are configured to process the chroma output features.

3. The method of claim 2, wherein the first quantity is equal to 24 and the second quantity is equal to 8, the first quantity is equal to 16 and the second quantity is equal to 8, or the first quantity is equal to 12 and the second quantity is equal to 4.

4. The method of claim 1, wherein the NN-based filter further comprises a transition block and an activation block, the transition block comprising one or more convolutions, the activation block being located after all convolutions in the one or more convolutions of the transition block.

5. The method of claim 1, wherein d6 ≥ sum (d1, d2, d3, d4, d4, d5), where d1 to d5 are the number of corresponding output channels, and d6 is the number of fused output channels of the single output.

6. The method of claim 5, wherein the plurality of input channels include indications of reconstructed samples, predicted samples, boundary intensity maps, second quantization parameters and first quantization parameters, and reconstructed samples, and wherein d1 = 4, d2 = 2, d3 = 1, d4 = 1, d5 = 1, and d6 = 16.

7. The method of claim 5, wherein the plurality of input channels include indications of reconstructed samples, predicted samples, boundary intensity maps, second quantization parameters and first quantization parameters, and wherein d1 = 12, d2 = 8, d3 = 4, d4 = 2, d5 = 2, and d6 = 32.

8. The method of claim 1, wherein the NN-based filter further comprises a transform block combined with a fusion block, wherein the fusion block comprises a 1×1 convolution block having a first number of channels (d6), wherein the transform block comprises a first separable convolution having a second number of channels (C1), a second separable convolution having a third number of channels (C2), and a third convolution having a fourth number of channels (C3), and wherein d6 = C1 = C2 = C3.

9. An apparatus for decoding video data, the apparatus comprising: One or more memories, the one or more memories being configured to store images of video data; and One or more processors, which communicate with one or more memories, are configured to: Receive video data indicating the image; Reconstructing blocks of the video data to generate reconstructed blocks; and 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 applying an NN-based filter, the NN-based filter including a restricted header block having multiple input channels and configured to independently extract a corresponding number of output channels for each of the multiple input channels and fuse the outputs associated with each of the multiple input channels into a single output, wherein the sum of the corresponding number of output channels is less than or equal to the number of fused output channels of the single output.

10. The device of claim 9, wherein the NN-based filter further comprises: A conversion block having a first number of luminance output features and a second number of chrominance output features, wherein the first number is greater than the second number; Multiple luminance backbone blocks, the multiple luminance backbone blocks being configured to process the luminance output features; and Multiple chroma backbone blocks are configured to process the chroma output features.

11. The device of claim 10, wherein the first quantity is equal to 24 and the second quantity is equal to 8, the first quantity is equal to 16 and the second quantity is equal to 8, or the first quantity is equal to 12 and the second quantity is equal to 4.

12. The device of claim 9, wherein the NN-based filter further comprises a transition block and an activation block, the transition block comprising one or more convolutions, the activation block being located after all convolutions in the one or more convolutions of the transition block.

13. The device of claim 9, wherein d6 ≥ sum (d1, d2, d3, d4, d4, d5), where d1 to d5 are the number of corresponding output channels, and d6 is the number of fused output channels of the single output.

14. The device of claim 13, wherein the plurality of input channels include indications of reconstructed samples, predicted samples, boundary intensity maps, second quantization parameters and first quantization parameters, and wherein d1 = 4, d2 = 2, d3 = 1, d4 = 1, d5 = 1, and d6 = 16.

15. The device of claim 13, wherein the plurality of input channels include indications of reconstructed samples, predicted samples, boundary intensity maps, second quantization parameters and first quantization parameters, and wherein d1 = 12, d2 = 8, d3 = 4, d4 = 2, d5 = 2, and d6 = 32.

16. The device of claim 9, wherein the NN-based filter further comprises a transform block combined with a fusion block, wherein the fusion block comprises a 1×1 convolution block having a first number of channels (d6), wherein the transform block comprises a first separable convolution having a second number of channels (C1), a second separable convolution having a third number of channels (C2), and a third convolution having a fourth number of channels (C3), and wherein d6 = C1 = C2 = C3.

17. The device of claim 9, wherein the device is configured to decode video data, and wherein the device further comprises a display configured to display the image.

18. An apparatus for encoding video data, the apparatus comprising: One or more memories, the one or more memories being configured to store images of video data; and One or more processors, which communicate with one or more memories, are configured to: Receive video data indicating the image; Reconstructing blocks of the video data to generate reconstructed blocks; and 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 applying an NN-based filter, the NN-based filter including a restricted header block having multiple input channels and configured to independently extract a corresponding number of output channels for each of the multiple input channels and fuse the outputs associated with each of the multiple input channels into a single output, wherein the sum of the corresponding number of output channels is less than or equal to the number of fused output channels of the single output.

19. The device of claim 18, wherein the NN-based filter further comprises: A conversion block having a first number of luminance output features and a second number of chrominance output features, wherein the first number is greater than the second number; Multiple luminance backbone blocks, the multiple luminance backbone blocks being configured to process the luminance output features; and Multiple chroma backbone blocks are configured to process the chroma output features.

20. The device of claim 18, wherein the NN-based filter further comprises a transition block and an activation block, the transition block comprising one or more convolutions, the activation block being located after all convolutions in the one or more convolutions of the transition block.