Filtering based on joint component neural network during video coding

Abstract

An example device for decoding video data includes a memory configured to store video data and one or more processors implemented in circuitry and configured to: apply a down-sampling convolutional neural network layer to a first color component of a block of video data, the first color component of the block having a first size, wherein applying the down-sampling convolutional neural network layer to the first color component generates a down-sampled first color component having a second size that is smaller than the first size; filter a second color component having the second size to form a filtered second color component; concatenate the down-sampled first color component and the filtered second color component to form a concatenated color component; and filter the concatenated color component to form a filtered concatenated component, the filtered concatenated component including a filtered down-sampled first color component.

Description

[0001] This application claims priority to U.S. Patent Application No. 17 / 493,543, filed October 4, 2021, and U.S. Provisional Application No. 63 / 087,784, filed October 5, 2020, the entire contents of each of which are incorporated herein by reference. U.S. Patent Application No. 17 / 493,543, filed October 4, 2021, claims the benefit of U.S. Provisional Application No. 63 / 087,784, filed October 5, 2020. Technical Field

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

[0003] Digital video capabilities can be incorporated into a wide range of devices, including digital televisions, digital direct-connect broadcasting 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 radio phones, so-called "smartphones," video conferencing equipment, video streaming devices, and more. Digital video devices implement video decoding technologies such as 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), and extensions thereof as defined in the standards described. By implementing such video decoding technologies, video devices can more efficiently transmit, 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 eliminate inherent redundancy in video sequences. For block-based video decoding, video slices (e.g., video pictures or portions of video pictures) can be segmented into video blocks, which may also be referred to as decoding tree units (CTUs), decoding units (CUs), and / or decoding nodes. Video blocks in intra-frame decoded (I) slices of a picture are encoded using spatial predictions with reference samples of adjacent blocks within the same picture. Video blocks in inter-frame decoded (P or B) slices of a picture can be encoded using spatial predictions with reference samples of adjacent blocks within the same picture or temporal predictions with reference samples of 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] Generally, this disclosure describes techniques for filtering decoded images, which may be distorted. The filtering process can be based on neural network techniques. The filtering process can be used in advanced video codecs, such as extensions to ITU-T TH.266 / Voice Universal Decoding (VVC), or subsequent generations of video decoding standards, as well as any other video codecs.

[0006] In one example, a method for filtering decoded video data includes: applying a downsampled convolutional neural network layer to a first color component of a block of video data, the first color component having a first size, wherein applying the downsampled convolutional neural network layer to the first color component generates a downsampled first color component having a second size smaller than the first size; filtering the second color component having the second size to form a filtered second color component; concatenating the downsampled first color component and the filtered second color component to form a concatenated color component; and filtering the concatenated color component to form a filtered concatenated component, the filtered concatenated component including the filtered downsampled first color component.

[0007] In another example, an apparatus for filtering decoded video data includes: a memory configured to store the video data; and one or more processors implemented in a circuit and configured to: apply a downsampled convolutional neural network layer to a first color component of a block of video data, the first color component having a first size, wherein applying the downsampled convolutional neural network layer to the first color component generates a downsampled first color component having a second size smaller than the first size; filter the second color component having the second size to form a filtered second color component; cascade the downsampled first color component and the filtered second color component to form a cascaded color component; and filter the cascaded color component to form a filtered cascaded component, the filtered cascaded component including the filtered downsampled first color component.

[0008] In another example, a computer-readable storage medium has instructions stored thereon that, when executed, cause a processor to: apply a downsampled convolutional neural network layer to a first color component of a block of video data, the first color component having a first size, wherein applying the downsampled convolutional neural network layer to the first color component generates a downsampled first color component having a second size smaller than the first size; filter the second color component having the second size to form a filtered second color component; concatenate the downsampled first color component and the filtered second color component to form a concatenated color component; and filter the concatenated color component to form a filtered concatenated component, the filtered concatenated component including the filtered downsampled first color component.

[0009] In another example, an apparatus for filtering decoded video data includes: means for applying a downsampled convolutional neural network layer to a first color component of a block of video data, the first color component having a first size, wherein applying the downsampled convolutional neural network layer to the first color component generates a downsampled first color component having a second size smaller than the first size; means for filtering the second color component having the second size to form a filtered second color component; means for concatenating the downsampled first color component and the filtered second color component to form a concatenated color component; and means for filtering the concatenated color components to form a filtered concatenated component, the filtered concatenated component including the filtered downsampled first color component.

[0010] Details of one or more examples are set forth in the accompanying drawings and description. Other features, objects, and advantages will become apparent from the specification, drawings, and claims. Attached Figure Description

[0011] Figure 1 This is a block diagram illustrating an example of a video encoding and decoding system that can perform the techniques disclosed herein.

[0012] Figure 2 This is a conceptual diagram illustrating a hybrid video decoding framework.

[0013] Figure 3 This is a conceptual diagram illustrating a hierarchical prediction structure using a group of pictures (GOP) of size 16.

[0014] Figure 4A and 4B This is a conceptual diagram illustrating an example filter based on a Joint Component Convolutional Neural Network (CNN) with a Residual Network (ResNet) that includes residual blocks.

[0015] Figure 5A and 5B This is a conceptual diagram showing an example quadtree binary tree (QTBT) structure and the corresponding decoding tree unit (CTU).

[0016] Figure 6 This is a block diagram illustrating an example video encoder that can perform the techniques disclosed herein.

[0017] Figure 7 This is a block diagram illustrating an example video decoder that can perform the techniques disclosed herein.

[0018] Figure 8 This is a block diagram illustrating an example of a joint component CNN filter according to the techniques of this disclosure.

[0019] Figure 9 This shows a conceptual diagram of an example residual processing unit.

[0020] Figure 10 This is a block diagram illustrating another example of a joint component CNN filter design according to the techniques of this disclosure.

[0021] Figure 11 This is a flowchart illustrating an example method for filtering a decoded block according to the techniques of this disclosure.

[0022] Figure 12 This is a flowchart illustrating an example method for filtering a decoded block according to the techniques of this disclosure.

[0023] Figure 13 This is a flowchart illustrating an example method for filtering decoded video data according to the techniques of this disclosure.

[0024] Figure 14 This is a flowchart illustrating an example method for filtering decoded video data according to the techniques of this disclosure. Detailed Implementation

[0025] 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 Coding (HEVC), or ITU-T H.265, including its Range Extension, Multiple View Extension (MV-HEVC), and Scalable Extension (SHVC). Another example video decoding standard is Multifunctional Video Coding (VVC), or ITU-TH.266, which was developed by the Joint Video Experts Team (JVET) of the ITU-T Video Coding Experts Group (VCEG) and the ISO / IEC Moving Picture Experts Group (MPEG). The first version of the VVC specification, hereinafter referred to as "VVC FDIS", can be obtained from http: / / phenix.int-evry.fr / jvet / doc_end_user / documents / 19_Teleconference / wg11 / JJVE-S2001-v17.zip.

[0026] Figure 1This is a block diagram illustrating an example video encoding and decoding system 100 that can perform the techniques of this disclosure. The techniques of this disclosure are generally directed to the decoding (encoding and / or decoding) of video data. Typically, video data includes any data used for processing video. Thus, video data can include raw, undecoded video, encoded video, decoded (e.g., reconstructed) video, and video metadata, such as signaling data.

[0027] like Figure 1 As illustrated in this example, system 100 includes source device 102, which provides encoded video data to be decoded and displayed by destination device 116. Specifically, source device 102 provides the video data to destination device 116 via computer-readable medium 110. Source device 102 and destination device 116 can include any of a variety of devices, including desktop computers, laptops, mobile devices, tablets, set-top boxes, telephones such as smartphones, televisions, cameras, display devices, digital media players, video game consoles, video streaming devices, and so on. In some cases, source device 102 and destination device 116 may be equipped for wireless communication and may therefore be referred to as wireless communication devices.

[0028] exist Figure 1 In the example, source device 102 includes a video source 104, a memory 106, a video encoder 200, and an output interface 108. Destination device 116 includes an input interface 122, a video decoder 300, a memory 120, and a 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 filtering techniques to the video data using a joint component neural network-based filtering process. 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, rather than including an integrated display device.

[0029] Figure 1The system 100 shown is merely an example. Typically, any digital video encoding and / or decoding device can perform filtering techniques on video data using filtering processes based on joint component neural networks. 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 (encoding and / or decoding) of data. Thus, video encoder 200 and video decoder 300 represent examples of decoding devices (specifically, video encoder and video decoder). In some examples, source device 102 and destination device 116 can operate in a substantially symmetrical manner, such that each of source device 102 and destination device 116 includes video encoding and decoding components. Therefore, system 100 can support one-way or two-way video transmission between source device 102 and destination device 116, for example, for video streaming, video playback, video broadcasting, or video telephony.

[0030] Typically, video source 104 represents the source of video data (i.e., raw, undecoded video data) and provides a continuous series of consecutive pictures (also referred to as "frames") of video data to video encoder 200, which encodes the picture data. Video source 104 of source device 102 may include video capture devices such as video 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 decode the pictures by rearranging them from the order of reception (sometimes referred to as "display order") to the decoding order. Video encoder 200 may generate a bitstream comprising the encoded video data. Then, source device 102 can output encoded video data to computer-readable medium 110 via output interface 108 for reception and / or retrieval by input interface 122 of destination device 116, for example.

[0031] 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 separate 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 data 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.

[0032] Computer-readable medium 110 can 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 transmit encoded video data directly to destination device 116 in real time (e.g., via a radio frequency network or a computer-based network). Output interface 108 can modulate the transmitted signal including the encoded video data, while input interface 122 can demodulate the received transmitted signal according to communication standards such as wireless communication protocols. The communication medium can include any wireless or wired communication medium, such as radio frequency (RF) spectrum or one or more physical transmission lines. The communication medium can form part of a packet-based network, such as a local area network, a wide area network, or a global network such as the Internet. The communication medium can include a router, a switch, a base station, or any other device that can be used to facilitate communication from source device 102 to destination device 116.

[0033] In some examples, source device 102 can output encoded data to storage device 112 from output interface 108. Similarly, destination device 116 can access encoded data from storage device 112 via input interface 122. Storage device 112 can include any of a variety of distributed or locally accessed data storage media, such as hard disks, 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.

[0034] In some examples, source device 102 may output encoded video data to file server 114 or another intermediate storage device that may store the encoded video data generated by source device 102. Destination device 116 may access the stored video data from file server 114 via streaming or download.

[0035] File server 114 can be any type of server device capable of storing encoded video data and sending such encoded video data to destination device 116. File server 114 can represent a web server (e.g., a website), a server configured to provide file transfer protocol services (e.g., File Transfer Protocol (FTP) or One-Way Transfer File Delivery (FLUTE) 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 Dynamic Adaptive Streaming (DASH), HTTP Real-Time Streaming (HLS), Real-Time Streaming Protocol (RTSP), HTTP Dynamic Streaming, or similar protocols.

[0036] 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 one 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.

[0037] Output interface 108 and input interface 122 may represent a wireless transmitter / receiver, a modem, a wired network 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 may be configured to transmit data such as encoded video data according to cellular communication standards (e.g., 4G, 4G-LTE (Long Term Evolution), LTE Advanced, 5G, or similar standards). In some examples where output interface 108 includes a wireless transmitter, output interface 108 and input interface 122 may be configured to operate according to other wireless standards (e.g., the IEEE 802.11 specification, the IEEE 802.15 specification (e.g., ZigBee)). TM ),Bluetooth TM The 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 to perform functions belonging to video encoder 200 and / or output interface 108, and destination device 116 may include an SoC device to perform functions belonging to video decoder 300 and / or input interface 122.

[0038] The technology disclosed herein can be applied to video decoding to support any of a variety of multimedia applications, such as over-the-air television broadcasting, cable television transmission, satellite television transmission, internet streaming video transmission such as HTTP Dynamic Adaptive Streaming (DASH), digital video encoded onto a data storage medium, decoding digital video stored on a data storage medium, or other applications.

[0039] 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 (such as syntax elements having values ​​describing the characteristics and / or processing of video blocks or other decoding units (e.g., slices, pictures, picture groups, sequences, etc.), which is also used by the video decoder 300. The display device 118 displays decoded images 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.

[0040] Although Figure 1Not shown, but in some examples, the video encoder 200 and video decoder 300 may each be integrated with the audio encoder and / or audio decoder, and may include appropriate MUX-DEMUX units, or other hardware and / or software, to handle multiplexed streams that include both audio and video in a common data stream. Where applicable, the MUX-DEMUX unit may conform to the ITU H.223 multiplexer protocol, or other protocols such as User Datagram Protocol (UDP).

[0041] The video encoder 200 and video decoder 300 can each be implemented as any of a variety of suitable encoder and / or decoder circuits, such as one or more microprocessors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), discrete logic, software, hardware, firmware, or any combination thereof. When these technologies are implemented in part in software, the device may store software instructions in a suitable non-transitory computer-readable medium and execute the instructions in hardware using one or more processors to perform the technologies of this disclosure. Each of the video encoder 200 and video decoder 300 may be included in one or more encoders or decoders, and either one may be integrated as part of a combined encoder / decoder (CODEC) in the respective device. Devices including the video encoder 200 and / or video decoder 300 may include integrated circuits, microprocessors, and / or wireless communication devices such as cellular phones.

[0042] The video encoder 200 and video decoder 300 may operate according to a video decoding standard (e.g., ITU-T H.265, also known as High Efficiency Video Decoding (HEVC)) or its extensions (e.g., Multi-View and / or Scalable Video Coding Extensions). Alternatively, the video encoder 200 and video decoder 300 may operate according to other proprietary or industry standards (e.g., Multi-Functional Video Coding (VVC)). A draft of the VVC standard is described in “Multi-Functional Video Coding (Draft 9)” by Bross et al., ITU-T SG 16WP 3 and the Joint Video Experts Team (JVET) of ISO / IEC JTC 1 / SC 29 / WG 11, 18th Meeting, 15-24 Apr., JVET-R2001-v8 (hereinafter referred to as “VVC Draft 9”). However, the techniques disclosed herein are not limited to any particular decoding standard.

[0043] Typically, video encoder 200 and video decoder 300 can perform block-based image decoding. The term "block" generally refers to a structure that includes data to be processed (e.g., encoded, decoded, or otherwise used for encoding and / or decoding processing). For example, a block may include a two-dimensional matrix of samples of luminance and / or chrominance data. Typically, video encoder 200 and video decoder 300 can decode video data represented in YUV (e.g., Y, Cb, Cr) format. That is, video encoder 200 and video decoder 300 can decode luminance and chrominance components, rather than decoding samples of red, green, and blue (RGB) data of an image, where chrominance components may include both red and blue hue chrominance components. In some examples, video encoder 200 converts the received RGB format data to a YUV representation before encoding, and video decoder 300 converts the YUV representation to RGB format. Alternatively, preprocessing and post-processing units (not shown) can perform these conversions.

[0044] This disclosure can generally relate to decoding (e.g., encoding and decoding) an image to include processing that encodes or decodes the image's data. Similarly, this disclosure can relate to decoding blocks of an image to include processing that encodes or decodes the block's data (e.g., prediction and / or residual decoding). Encoded video bitstreams typically include a series of values ​​for representing decoding decisions (e.g., decoding modes) and syntax elements that segment the image into blocks. Therefore, references to decoding an image or block should generally be understood as decoding the values ​​of the syntax elements that form the image or block.

[0045] HEVC defines various blocks, including decoding units (CUs), prediction units (PUs), and transform units (TUs). According to HEVC, a video encoder (e.g., video encoder 200) partitions a decoding tree unit (CTU) into CUs based on a quadtree structure. That is, the video decoder partitions 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 can be called "leaf nodes," and the CU of such leaf nodes can include one or more PUs and / or one or more TUs. The video decoder can further partition PUs and TUs. For example, in HEVC, a residual quadtree (RQT) represents a partition of a TU. In HEVC, PUs represent inter-frame prediction data, while TUs represent residual data. Intra-frame prediction CUs include intra-frame prediction information such as intra-frame mode indications.

[0046] As another example, video encoder 200 and video decoder 300 can be configured to operate according to VVC. According to VVC, the video encoder (e.g., video encoder 200) segments the image into multiple decoding tree units (CTUs). Video encoder 200 can segment CTUs according to a tree structure (e.g., a quadtree-binary tree (QTBT) structure or a multi-type tree (MTT) structure). The QTBT structure removes the concept of multiple segmentation types (e.g., the separation between CUs, PUs, and TUs in HEVC). The QTBT structure includes two levels: a first-level segmentation based on quadtree segmentation and a second-level segmentation based on binary tree segmentation. The root node of the QTBT structure corresponds to a CTU. The leaf nodes of the binary tree correspond to decoding units (CUs).

[0047] In the MTT partitioning structure, blocks can be partitioned using quadtree (QT) partitioning, binary tree (BT) partitioning, and one or more types of ternary tree (TT) partitioning (also known as tripartite tree (TT)) partitioning. A ternary tree or tripartite tree partitioning is a partition in which a block is split into three sub-blocks. In some examples, a ternary tree or tripartite tree partitioning divides a block into three sub-blocks without dividing the original block through a center. The partitioning types in MTT (e.g., QT, BT, and TT) can be symmetric or asymmetric.

[0048] 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. However, 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).

[0049] The video encoder 200 and video decoder 300 can be configured to use quadtree segmentation, QTBT segmentation, MTT segmentation, or other segmentation structures according to HEVC. For illustrative purposes, the description of the technology disclosed herein is presented with respect to QTBT segmentation. However, it should be understood that the technology disclosed herein can also be applied to video decoders configured to use quadtree segmentation or other types of segmentation.

[0050] In some examples, a CTU includes a coded tree block (CTB) of luma samples, two corresponding CTBs of chroma samples of an image with three sample arrays, or a CTB of samples of a monochrome image or an image using three independent color planes and a decoding syntax structure for the samples. A CTB can be an N×N block of samples for some value of N, such that dividing a component into a CTB is a segmentation. A component is an array or a single sample from one of three arrays (luma and two chroma) that make up the image in a 4:2:0, 4:2:2, or 4:4:4 color format; or a component is an array or a single sample of the image in a monochrome format. In some examples, a decoding block is an M×N block of samples for some values ​​of M and N, such that dividing a CTB into a decoding block is a segmentation.

[0051] 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 ​​CTUs with a height equal to the image's height and a width specified by a syntax element (e.g., in the image parameter set). A tile row refers to a rectangular area of ​​CTUs with a height specified by a syntax element (e.g., in the image parameter set) and a width equal to the image's width.

[0052] In some examples, a tile can be divided into multiple bricks, each brick 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, a brick that is a proper subset of a tile cannot be called a tile.

[0053] The bricks in an image can also be arranged as slices. A slice can be an integer number of bricks in the image, which can be exclusively contained within a single Network Abstraction Layer (NAL) unit. In some examples, a slice consists of multiple complete bricks or a continuous sequence of complete bricks that contain only one tile.

[0054] This disclosure uses "N×N" and "N by N" interchangeably to refer to the sample dimensions of a block (such as a CU or other video block) in the vertical and horizontal dimensions, for example, 16×16 samples or 16 by 16 samples. Typically, a 16×16 CU has 16 samples in the vertical direction (y = 16) and 16 samples in the horizontal direction (x = 16). Similarly, an N×N CU typically has N samples in the vertical direction and N samples in the horizontal direction, where N represents a non-negative integer value. Samples in a CU can be arranged in rows and columns. Furthermore, a CU does not necessarily have the same number of samples in the horizontal direction as it does in the vertical direction. For example, a CU can include N×M samples, where M is not necessarily equal to N.

[0055] The video encoder 200 encodes video data containing representations of CUs, prediction and / or residual information, as well as other information. Prediction information indicates how to predict CUs to form prediction blocks of CUs. Residual information typically represents the sample-by-sample difference between the prediction block and a sample of the CU before encoding.

[0056] To predict the Cubic Frame (CU), the video encoder 200 typically forms a prediction block of the CU through inter-frame prediction or intra-frame prediction. Inter-frame prediction generally refers to predicting the CU from previously decoded picture data, while intra-frame prediction generally refers to predicting the CU from previously decoded data of the same picture. To perform inter-frame prediction, the video encoder 200 can generate prediction blocks using one or more motion vectors. The video encoder 200 can typically perform motion search to identify, for example, a reference block that closely matches the CU in terms of the difference between the CU and a reference block. The video encoder 200 can compute difference metrics such as Sum of Absolute Differences (SAD), Sum of Squared Differences (SSD), Mean Absolute Difference (MAD), Mean Squared Difference (MSD), or other such difference calculations to determine whether the reference block closely matches the current CU. In some examples, the video encoder 200 can use unidirectional or bidirectional prediction to predict the current CU.

[0057] Some examples of VVC also provide affine motion compensation modes, which can be considered a type of 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).

[0058] 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. Typically, the video encoder 200 selects an intra-frame prediction mode that describes the prediction of samples for the current block from samples adjacent to those of the current block (e.g., a block of the CU). Assuming the video encoder 200 decodes the CTU and CU in raster scan order (from left to right, from top to bottom), such samples are typically located above, to the upper left, or to the left of the current block within the same frame as the current block.

[0059] 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 and the motion information of the corresponding mode. For example, for unidirectional or bidirectional inter-frame prediction, the video encoder 200 may use Advanced Motion Vector Prediction (AMVP) or merging modes to encode motion vectors. The video encoder 200 may use similar modes to encode motion vectors for affine motion compensation modes.

[0060] After block prediction (such as intra-frame prediction or inter-frame prediction), the video encoder 200 can compute residual data for that block. Residual data, such as that of a residual block, represents the sample-by-sample difference between that block and its predicted block 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 Discrete Cosine Transform (DCT), integer transform, wavelet transform, or conceptually similar transforms to the residual video data. Furthermore, the video encoder 200 can apply secondary transforms after the primary transform, such as Mode-Dependent Inseparable Secondary Transform (MDNSST), signal-dependent transforms, Karhunen-Loeve Transform (KLT), etc. The video encoder 200 produces transform coefficients after applying one or more transforms.

[0061] As described above, after performing any transformation to produce transform coefficients, the video encoder 200 can perform quantization of the transform coefficients. Quantization generally refers to quantizing the transform coefficients to potentially reduce the amount of data used to represent them, thereby providing further compression. By performing quantization, the video encoder 200 can reduce the bit depth associated with some or all of the transform coefficients. For example, the video encoder 200 can 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 can perform a bit-by-bit right shift of the value to be quantized.

[0062] After quantization, the video encoder 200 can scan the transform coefficients to generate a one-dimensional vector from a two-dimensional matrix including the quantized transform coefficients. The scan can be designed to place coefficients with higher energy (and therefore lower frequencies) before the vector and transform coefficients with lower energy (and therefore higher frequencies) 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 the 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 for use by the video decoder 300 in decoding the video data.

[0063] To perform CABAC, the video encoder 200 can assign context from a context model to the symbols to be transmitted. For example, the context might involve whether the neighboring values ​​of a symbol are zero. Probability determination can be based on the context assigned to the symbols.

[0064] The video encoder 200 can also generate syntax data for the video decoder 300, for example, in image headers, block headers, or slice headers. This syntax data includes block-based syntax data, image-based syntax data, sequence-based syntax data, and other syntax data (e.g., sequence parameter sets (SPS), image parameter sets (PPS), or video parameter sets (VPS)). The video decoder 300 can also decode this syntax data to determine how to decode the corresponding video data.

[0065] In this way, the video encoder 200 can generate a bitstream comprising encoded video data, such as syntax elements describing the segmentation of the image 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.

[0066] Typically, the video decoder 300 performs the inverse of the process performed by the video encoder 200 to decode the encoded video data of the bitstream. For example, the video decoder 300 can use CABAC to decode the values ​​of the syntax elements of the bitstream in a manner substantially similar to but inverse of the CABAC encoding process of the video encoder 200. Syntax elements can define segmentation information for dividing the image into CTUs and for segmenting each CTU according to a corresponding segmentation structure (e.g., a QTBT structure) to define the CUs of the CTUs. Syntax elements can also define prediction and residual information for blocks (e.g., CUs) of the video data.

[0067] 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 of the block. The video decoder 300 uses the prediction mode (intra-frame prediction or inter-frame prediction) and associated prediction information (e.g., motion information for inter-frame prediction) as instructed by signaling to form a prediction block of 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 deblocking, to reduce visual artifacts along the block edges.

[0068] This disclosure can generally refer to "signaling notification," such as certain information about syntax elements. The term "signaling notification" can generally refer to communication of values ​​for syntax elements and / or other data used to decode encoded video data. That is, the video encoder 200 can signal the values ​​of syntax elements in the bitstream. Typically, signaling notification refers to generating values ​​in the bitstream. As described above, the source device 102 can transmit the bitstream to the destination device 116 substantially in real time, or (for example, this may happen when syntax elements are stored in storage device 112 for later retrieval by the destination device 116) not in real time.

[0069] Figure 2 This is a conceptual diagram illustrating a hybrid video decoding framework. Video decoding standards since H.261 have all been based on the so-called hybrid video decoding principle, which... Figure 2 As shown in the diagram. The term hybrid refers to combining two methods to reduce redundancy in video signals: prediction and transform decoding with the quantization of the prediction residuals. Prediction and transform reduce redundancy in video signals through decorrelation, while quantization reduces the data represented by the transform coefficients by lowering their precision, ideally by removing only irrelevant details. This hybrid video decoding design principle has also been used in two recent standards, ITU-T H.265 / HEVC and ITU-TH.266 / VVC.

[0070] like Figure 2 As shown, a modern hybrid video decoder 130 typically includes block segmentation, 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.

[0071] Typically, the video decoder 130 receives input video data 132 when decoding video data. Block segmentation is an operation used to divide the received video data into smaller blocks for prediction and transform processing. Early video decoding standards used fixed block sizes, typically 16×16 samples. More recent standards, such as HEVC and VVC, employ tree-based segmentation structures to provide flexible segmentation.

[0072] Motion estimation unit 156 and inter-frame prediction unit 154 can predict the input video data 132, for example, from previously decoded data from DPB 150. Motion compensation, or inter-picture prediction, utilizes the redundancy present between pictures (and thus "inter-frames") in the video sequence. According to block-based motion compensation used in all modern video codecs, the prediction is derived from one or more previously decoded pictures (i.e., reference pictures). The corresponding region for generating inter-frame predictions is indicated by motion information, including motion vectors and reference picture indices.

[0073] The summing unit 134 can calculate 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.

[0074] 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 produce 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 produce a filtered decoded block.

[0075] Video data blocks (such as CTUs or CUs) can actually include multiple color components, such as a luma or “luminance” component, a blue-hue chroma or “chroma” component, and a red-hue chroma (chroma) component. A luma component can have a larger spatial resolution than a chroma component, and one chroma component can have a larger spatial resolution than the other. Alternatively, a luma component can have a larger spatial resolution than a chroma component, while two chroma components can have equal spatial resolution. For example, in a 4:2:2 format, the luma component can be twice the size of the chroma component in the horizontal direction and equal to the chroma component in the vertical direction. As another example, in a 4:2:0 format, the luma component can be twice the size of the chroma component in both the horizontal and vertical directions. The various operations discussed above can generally be applied individually to each of the luma and chroma components (although some decoding information, such as motion information or intra-frame prediction direction, can be determined for the luma component and inherited by the corresponding chroma component).

[0076] According to the technology disclosed herein, the loop filter unit 148 can receive from the summing unit 146 a first color component (e.g., a luminance or "luminance" component) having a first size and a second color component (e.g., a blue or red hue chroma or "chroma" component) having a second size smaller than the first size. A common video data block may include both the first and second color components. The loop filter unit 148 can be configured to apply a downsampled convolutional neural network layer to the first color component of the video data block to generate a downsampled first color component with a second size, i.e., downsample the first color component to the size of the second color component. In one example, the first color component may be a luminance component, and the second component may be one of two chroma components. In another example, the first color component may be a first chroma component, and the second color component may be a second chroma component.

[0077] In another example, the loop filter unit 148 may receive a luminance component and each of two chrominance components, wherein the luminance component has a first size, the first chrominance component has a second size smaller than the first size, and the second chrominance component has a third size smaller than the second size. In this example, the loop filter unit may apply a downsampled convolutional neural network filter to both the luminance component and the first chrominance component to generate a downsampled luminance component with a third size and a downsampled first chrominance component with a third size.

[0078] The loop filtering unit 148 can also filter the second color component to form a filtered second color component, for example, using a convolutional neural network filter. The loop filtering unit 148 can then cascade the downsampled first color component and the filtered second color component to form a cascaded color component. The loop filtering unit 148 can then filter the cascaded color component (e.g., using a convolutional neural network filter) to form a filtered cascaded component, wherein the filtered cascaded component includes the filtered downsampled first color component.

[0079] Specifically, the first and second color components may initially be stored in separate arrays or matrices. To cascade the color components, the loop filter unit 148 can form a single array or matrix with a width or height twice that of a single color component. The loop filter unit 148 can then store samples of the first color component in a first region of the newly formed array or matrix and samples of the second color component in an adjacent second region of the newly formed array or matrix. When using three color components (e.g., luminance, blue hue chroma, and red hue chroma), the loop filter unit 148 can form a single array or matrix with a width or height three times that of a single color component and store samples of each of the three color components in the corresponding region of the newly formed array or matrix.

[0080] As described above, the loop filter unit 148 can receive each of the luminance component and the chrominance component. The loop filter unit 148 can filter both the downsampled first chrominance component and the second chrominance component (e.g., using a convolutional neural network filter) to generate filtered downsampled first chrominance component and filtered second chrominance component. The loop filter unit 148 can then cascade the downsampled luminance component with the filtered downsampled first chrominance component and the filtered second chrominance component to form cascaded color components. The loop filter unit 148 can then filter the cascaded color components (including the downsampled luminance component, the filtered downsampled first chrominance component, and the filtered second chrominance component) (e.g., using a convolutional neural network filter).

[0081] After filtering the cascaded color components, the loop filter unit 148 can also upsample the filtered, downsampled first color component back to the first size (i.e., the original size of the first color component). If the loop filter unit 148 also downsamples the second color component (e.g., from the second size to the third size), it can also upsample the filtered second color component to the second size (i.e., the original size of the second color component).

[0082] Figure 3This is a conceptual diagram illustrating a hierarchical prediction structure 166 using picture groups (GOPs) of size 16. In recent video codecs, hierarchical prediction structures within picture groups (GOPs) have been applied to improve decoding efficiency.

[0083] Refer again Figure 2 Intra-frame prediction utilizes the spatial redundancy present in an image by deriving predictions for blocks from already encoded / decoded, spatially adjacent (reference) samples (hence the term "intra"). Modern video codecs (including AVC, HEVC, and VVC) employ directional angle prediction, DC prediction, and planar prediction.

[0084] Hybrid video decoding standards apply block transforms to the prediction residuals (whether they originate from predictions between or within images). Early standards, including H.261, H.262, and H.263, employed Discrete Cosine Transform (DCT). In HEVC and VVC, in addition to DCT, more transform kernels are applied to account for the different statistics present in specific video signals.

[0085] The purpose of quantization is to reduce the precision of the input values ​​or sets of input values, thereby reducing the amount of data required to represent those values. In hybrid video decoding, quantization is typically applied to the residual samples of a single transform, i.e., the transform coefficients, to obtain integer coefficient levels. In recent video decoding standards, the stride is derived from a so-called quantization parameter (QP) that controls fidelity and bit rate. A larger stride reduces the bit rate but also reduces quality, which, for example, causes video images to exhibit blocky artifacts and blurred details.

[0086] Context Adaptive Binary Arithmetic Coding (CABAC) is an entropy decoding form used in recent video codecs (e.g., AVC, HEVC, and VVC) due to its high efficiency.

[0087] Post-loop / in-loop filtering is a filtering process (or a combination of such processes) applied to reconstructed images to reduce coded artifacts. The input to the filtering process is typically the reconstructed image, which is a combination of the reconstructed residual signal (including quantization errors) and predictions. For example... Figure 2 As shown, the reconstructed image after in-loop filtering is stored and used as a reference for inter-image prediction of subsequent images. Decoding artifacts are primarily determined by the QP (Quick Pointer), therefore QP information is typically 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, the Adaptive Loop Filter (ALF) is introduced as a third filter. The ALF filtering process is shown below:

[0088] R′(i,j)=R(i,j)+((∑ k≠0 ∑ l≠0f(k,l)×K(R(i+k,j+l)-R(i,j),c(k,l))+64)>>7) (1)

[0089] Where R(i,j) is the sample set before filtering. R′(i,j) is the sample value after filtering. f(k,l) represents the filtering coefficients. K(x,y) is the clipping function, and c(k,l) represents the clipping parameters. Variables k and l are used in... and The length of the filter varies, where L represents the length of the filter. The clipping function K(x, y) = min(y, max(-y, x)), which corresponds to the function Clip3(-y, y, x). Clipping introduces nonlinearity, making ALF more efficient by reducing the influence of neighboring sample values ​​that differ too much from the current sample value. In VVC, the filter parameters can be signaled in the bitstream, and the filter parameters can be selected from a predefined set of filters. The filtering process of ALF can also be summarized by the following formula:

[0090] R'(i,j)=R(i,j)+ALF_residual_ouput(R) (2)

[0091] Figure 4A and 4B This is a conceptual diagram illustrating an example filter based on a Joint Component Convolutional Neural Network (CNN) with a Residual Network (ResNet) including residual blocks. Specifically, Figure 4A System 170 is depicted, which includes a convolutional neural network (conv.NN) filter 172, a cascaded unit 174, a series of residual blocks 176, residual blocks 178A, 178B (residual block 178), and channel-wise adder units 179A, 179B (channel-wise adder unit 179). Figure 4B Example residual block 180 is depicted. Any of residual blocks 176 and 178 may include components similar to residual block 180. Figure 4B In the example, residual block 180 includes convolutional NN filters 182 and 186, a rectified linear unit (ReLU) layer 184, and a summing unit 188.

[0092] exist Figure 4A In the example, convolutional NN filter 172 upsamples the chroma components (i.e., blue hue (Cb) and red hue (Cr) components) to the corresponding luminance (luminance (Y) component) size. Cascade unit 174 concatenates the upsampled chroma and luminance samples as input to the NN-based filter formed by residual block 176. The filtering process performed by residual block 176 can also be summarized as follows:

[0093] R'(i,j)=R(i,j)+NN_filter_residual_ouput(R) (3)

[0094] The loop filters of the video encoder 200 and the video decoder 300 may include similar to Figure 4A The system 170 is a component of the NN-based filter. The model structure and model parameters of the NN-based filter can be predefined and stored at the video encoder 200 and the video decoder 300. Additionally or alternatively, the video encoder 200 can signal the data signal representing the NN-based filter in the bitstream.

[0095] After residual block 176 applies NN-based filters to the cascaded components, residual blocks 178A and 178B can apply channel-specific filters to individual color components, and then channel-by-channel adder units 179A and 179B can separate the components to their respective color channels Y, Cb, and Cr.

[0096] exist Figure 4B In the example, convolutional NN filter 182 can first filter samples of the received video data. Convolutional NN filter 182 and convolutional NN filter 186 can be 3×3×K×K convolutional NN filters. Residual block 180 can also be referred to as a "residual processing unit". ReLU layer 184 can provide an activation function that, for each sample, outputs the sample value itself when the sample value is greater than zero, otherwise, if the input sample value is less than zero, it outputs zero for that sample value. In other words, the ReLU activation function can be summarized as follows:

[0097]

[0098] Then, the convolutional neural network filter 186 can filter the result sample values ​​from the ReLU layer 184 and provide the filtered sample values ​​to the summing unit 188. The summing unit 188 can add the filtered sample values ​​to the corresponding input sample values.

[0099] This disclosure recognizes that, in some cases, the size of the neural network may increase and the computational complexity may also increase as a result of upsampling chroma blocks to the corresponding luminance blocks.

[0100] Figure 1 Video encoder 200 and video decoder 300 or Figure 2The video decoder 130 can be configured, according to the techniques disclosed herein, to perform any or all of the following techniques individually or in any combination. Typically, the video encoder 200, video decoder 300, and video decoder 130 can first apply convolutional neural network (CNN) layers individually to some or all of the color components to align sizes among the color components. The video encoder 200 and video decoder 300 can set the size of the output neural (or tensor) block to be equal to the size of the color component block with the smallest size among all color components. The video encoder 200 and video decoder 300 can then apply (multiple) joint convolutional NN layers to all components with the same size. As a specialized and general application, for a video sequence in 4:2:0 color format, the luminance component block is downsampled using a convolutional NN layer with a stride of 2 in both the horizontal and vertical directions. The output of the convolutional NN layer can have the same block size as the chrominance component block size. The video encoder 200 and video decoder 300 can then apply the data from all color components to the joint convolutional NN layer as input.

[0101] Figure 5A and 5B This is a conceptual diagram illustrating an example Quadtree Binary Tree (QTBT) structure 190 and its corresponding Coding Tree Unit (CTU) 192. Solid lines represent quadtree splits, and dashed lines represent binary tree splits. In each split (i.e., non-leaf) node of the binary tree, a signaling flag indicates which split type (i.e., horizontal or vertical) is used, where 0 indicates a horizontal split and 1 indicates a vertical split in this example. For quadtree splits, it is not necessary to indicate the split type because the quadtree node splits the block horizontally and vertically into four equal-sized sub-blocks. Therefore, the video encoder 200 can encode syntax elements (e.g., split information) for the region tree level (i.e., solid lines) of the QTBT structure 190 and syntax elements (e.g., split information) for the prediction tree level (i.e., dashed lines) of the QTBT structure 190, and the video decoder 300 can decode them. The video encoder 200 can encode video data (such as prediction and transform data) of the CU represented by the terminal leaf nodes of the QTBT structure 190, and the video decoder 300 can decode it.

[0102] Typically, Figure 5BThe CTU 192 can be associated with parameters that define the size of the blocks corresponding to the nodes at the first and second levels of the QTBT structure 190. These parameters can include the CTU size (representing the size of the CTU 192 in the sample), the minimum quadtree size (MinQTSize, representing the minimum allowed quadtree leaf node size), the maximum binary tree size (MaxBTSize, representing the maximum allowed binary tree root node size), the maximum binary tree depth (MaxBTDepth, representing the maximum allowed binary tree depth), and the minimum binary tree size (MinBTSize, representing the minimum allowed binary tree leaf node size).

[0103] The root node of the QTBT structure corresponding to CTU can have four child nodes at the first level of the QTBT structure, and each child node can be partitioned according to the quadtree split. That is, the nodes at the first level are either leaf nodes (with no child nodes) or have four child nodes. An example of QTBT structure 190 represents such a node as including a parent node and child nodes with solid lines for branching. If the nodes at the first level are not larger than the maximum allowed binary tree root node size (MaxBTSize), these nodes can be further partitioned by the corresponding binary tree. The binary tree split of a node can be iterated until the nodes resulting from the split reach the minimum allowed binary tree leaf node size (MinBTSize) or the maximum allowed binary tree depth (MaxBTDepth). An example of QTBT structure 190 represents such nodes as having dashed lines for branching. The leaf nodes of the binary tree are called decoding units (CUs), which are used for prediction (e.g., intra-picture or inter-picture prediction) and transformation without any further splitting. As discussed above, CUs can also be referred to as “video chunks” or “blocks”.

[0104] In one example of a QTBT segmentation structure, the CTU size is set to 128×128 (luminance sample and two corresponding 64×64 chrominance samples), MinQTSize is set to 16×16, MaxBTSize is set to 64×64, MinBTSize (width and height) is set to 4, and MaxBTDepth is set to 4. First, quadtree segmentation is applied to the CTU to generate quadtree leaf nodes. Quadtree leaf nodes can have sizes ranging from 16×16 (i.e., MinQTSize) to 128×128 (i.e., the CTU size). If a quadtree leaf node is 128×128, it will not be further split by the binary tree because its size exceeds MaxBTSize (i.e., 64×64 in this example). Otherwise, the quadtree leaf node can be further split by the binary tree. Therefore, the quadtree leaf node is also the root node of the binary tree and has a binary tree depth of 0. When the binary tree depth reaches MaxBTDepth (4 in this example), further splitting is not allowed. A binary tree node with a width equal to MinBTSize (4 in this example) means that further vertical splits (i.e., width partitions) are not allowed for that node. Similarly, a binary tree node with a height equal to MinBTSize means that further horizontal splits (i.e., height partitions) are not allowed for that node. As mentioned above, the leaf nodes of the binary tree are called CUs and are further processed according to predictions and transformations without further splitting.

[0105] Figure 6 This is a block diagram illustrating an example video encoder 200 that can perform the techniques of this disclosure. Figure 6 This disclosure is provided for illustrative purposes and should not be construed as limiting the techniques illustrated and described herein. For illustrative purposes, this disclosure describes the video encoder 200 in the context of video coding standards, such as the ITU-T TH.265 / HEVC video coding standard and the VVC video coding standard under development. However, the techniques disclosed herein are not limited to these video coding standards and are generally applicable to other video coding and decoding standards.

[0106] exist Figure 6In 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 encoding 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 encoding unit 220 can be implemented in one or more processors or processing circuits. For example, the units of the video encoder 200 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 encoder 200 may include additional or alternative processors or processing circuits to perform these and other functions.

[0107] The video data storage device 230 can store video data to be encoded by the components of the video encoder 200. The video encoder 200 can obtain data from, for example, a video source 104 (…). Figure 1 The video encoder 200 receives video data stored in video data memory 230. DPB 218 can act as a reference picture memory, storing reference video data for use by the video encoder 200 when predicting subsequent video data. 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. Video data memory 230 and DPB 218 can be provided by the same memory device or separate memory devices. In different examples, as shown, video data memory 230 can be on-chip with other components of the video encoder 200, or off-chip relative to these components.

[0108] In this disclosure, references to video data memory 230 should not be construed as being limited to memory within video encoder 200 unless specifically described as such, or to memory external to video encoder 200 unless specifically described as such. Rather, references to video data memory 230 should be understood as references to memory that stores video encoder 200 receiving video data for encoding (e.g., video data of the current block to be encoded). Figure 1 The memory 106 can also provide temporary storage for the outputs from the various units of the video encoder 200.

[0109] Figure 6Various units are shown to aid in understanding the operations performed by the video encoder 200. These units can be implemented as fixed-function circuits, programmable circuits, or a combination thereof. Fixed-function circuits refer to circuits that provide a specific function and pre-configured the operable operations. Programmable circuits refer to circuits that can be programmed to perform various tasks and provide flexible functionality in the operable operations. For example, a programmable circuit can execute software or firmware such that the programmable circuit operates in a manner defined by instructions from 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 of the units may be different circuit blocks (fixed-function or programmable), and in some examples, one or more units may be integrated circuits.

[0110] 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., object 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.

[0111] 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 the raw video data to be encoded.

[0112] 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 to perform 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.

[0113] The mode selection unit 202 typically coordinates multiple coding passes to test combinations of coding parameters and the rate-distortion values ​​resulting from these combinations. Coding parameters may include the CTU-CU split, the prediction mode for the CU, the transformation type of the residual data for the CU, and the quantization parameters of the residual data for the CU. The mode selection unit 202 can ultimately select a combination of coding parameters that has a better rate-distortion value than other tested combinations.

[0114] The video encoder 200 can segment images 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 segment the CTUs of the image according to a tree structure (such as the QTBT structure described above or the quadtree structure of HEVC). As described above, the video encoder 200 can form one or more CUs from the segmented CTUs according to the tree structure. Such CUs can also be commonly referred to as "video blocks" or "blocks".

[0115] Typically, 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 predicted blocks for the current block (e.g., the current CU, or the overlapping portion of PU and TU in HEVC). For inter-frame prediction of the current block, motion estimation unit 222 may perform a motion search to identify one or more closely matching reference blocks in reference images (e.g., one or more previously decoded images 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 the sample-by-sample difference between the current block and the reference block being considered. Motion estimation unit 222 may identify the reference block with the lowest value derived from these calculations, thereby indicating the reference block that most closely matches the current block.

[0116] Motion estimation unit 222 can generate one or more motion vectors (MVs) that define the position of a reference block in a reference image relative to a current block in the current image. Motion estimation unit 222 can then provide these 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, it can provide two motion vectors. Motion compensation unit 224 can then use the motion vectors to generate prediction blocks. For example, motion compensation unit 224 can use the motion vectors to retrieve data for reference blocks. As another example, if the motion vectors have fractional-sample accuracy, motion compensation unit 224 can interpolate the values ​​of the prediction blocks 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 by, for example, sample-by-sample averaging or weighted averaging.

[0117] As another example, for intra-prediction or intra-prediction decoding, intra-prediction unit 226 can generate a prediction block from samples adjacent to the current block. For example, in directional mode, intra-prediction unit 226 can typically mathematically combine the values ​​of adjacent samples and fill these calculated values ​​in a defined direction across the current block to produce a prediction block. As another example, in DC mode, intra-prediction unit 226 can calculate the average of samples adjacent to the current block and generate a prediction block to include this average for each sample of the resulting prediction block.

[0118] Mode selection unit 202 provides the prediction block to residual generation unit 204. Residual generation unit 204 receives the raw, uncoded version of the current block from video data memory 230 and the prediction block from mode selection unit 202. Residual generation unit 204 calculates the sample-by-sample difference between the current block and the prediction block. The resulting sample-by-sample difference defines the residual block of the current block. In some examples, residual generation unit 204 may also determine the difference between sample values ​​in the residual block to generate the residual block using Residual Pulse Decode Modulation (RDPCM). In some examples, residual generation unit 204 may be formed using one or more subtractor circuits that perform binary subtraction.

[0119] 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 indicated above, the size of a CU can refer to the size of the luma decoding block of the CU, while the size of a PU can refer to the size of the luma prediction unit of the PU. Assuming a particular CU has a size of 2N×2N, video encoder 200 can support PU sizes of 2N×2N or N×N for intra-frame prediction, and symmetrical PU sizes of 2N×2N, 2N×N, N×2N, N×N, or similar sizes for inter-frame prediction. Video encoder 200 and video decoder 300 can also support asymmetric segmentation of PU sizes of 2N×nU, 2N×nD, nL×2N, and nR×2N for inter-frame prediction.

[0120] In the 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.

[0121] For other video decoding techniques, such as intra-block copy mode decoding, affine mode decoding, and linear model (LM) mode decoding, the 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, the mode selection unit 202 may not generate a prediction block, but instead may generate syntax elements indicating how the block can be reconstructed based on the selected palette. In this mode, the mode selection unit 202 may provide these syntax elements to the entropy coding unit 220 for encoding.

[0122] As described above, the residual generation unit 204 receives video data of 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.

[0123] Transform processing unit 206 applies one or more transformations to the residual block to produce 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 discrete cosine transform (DCT), direction transformation, Karl Henroff 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, such as rotation transformations. In some examples, transform processing unit 206 does not apply any transformations to the residual block.

[0124] Quantization unit 208 can quantize the transform coefficients in a transform coefficient block to produce a quantized transform coefficient block. Quantization unit 208 can quantize the transform coefficients of the transform coefficient block based on 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 introduce information loss; therefore, the quantized transform coefficients may have lower precision than the original transform coefficients generated by transform processing unit 206.

[0125] The inverse quantization unit 210 and the inverse transform processing unit 212 can apply inverse quantization and inverse transform to the quantized transform coefficient block, respectively, to reconstruct the residual block from the transform coefficient block. The reconstruction unit 214 can generate a reconstructed block corresponding to the current block based on the reconstructed residual block and the prediction block generated by the mode selection unit 202 (although there may be some distortion). 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.

[0126] Filter unit 216 can perform one or more filtering operations on the reconstructed block. For example, filter unit 216 can perform a deblocking operation to reduce block artifacts along the edges of the CU. In some examples, the operations of filter unit 216 can be skipped. Filter unit 216 can be configured to perform various techniques of this disclosure, such as resizing the color components to the smallest possible color component size and then applying a convolutional neural network (CNN) filter 232 to the resized color components.

[0127] Specifically, filter unit 216 can apply one or more CNN filters 232 to reduce the size of the luminance component to the size of a smaller chrominance component in the chrominance components, instead of upsampling the chrominance component to the size of the corresponding luminance component. That is, filter unit 216 can receive a luminance component with a first size and a chrominance component with a second size smaller than the first size. Filter unit 216 can apply a convolutional NN filter of the CNN filter 232 to the luminance component to downsample the luminance component to the second size. Filter unit 216 can filter the chrominance component to form a filtered chrominance component. Filter unit 216 can then cascade the downsampled luminance component and the filtered chrominance component to form a cascaded color component, and then filter the cascaded color component to form a filtered cascaded component, which includes the filtered downsampled first color component. Furthermore, filter unit 216 can upsample the filtered luminance component of the second size back to the first size before storing the luminance component in DPB 218.

[0128] Specifically, the first and second color components can initially be stored in separate arrays or matrices. To cascade the color components, filter unit 216 can form a single array or matrix with a width or height twice that of a single color component. Filter unit 216 can then store samples of the first color component in a first region of the newly formed array or matrix and samples of the second color component in an adjacent second region of the newly formed array or matrix. When using three color components (e.g., luminance, blue hue chroma, and red hue chroma), filter unit 216 can form a single array or matrix with a width or height three times that of a single color component and store samples of each of the three color components in the corresponding region of the newly formed array or matrix.

[0129] The video encoder 200 stores reconstructed blocks in the DPB 218. For example, in an example where the operation of the filter unit 216 is not required, the reconstruction unit 214 can store the reconstructed blocks in the DPB 218. In an example where the operation of the filter unit 216 is required, the filter unit 216 can store filtered reconstructed blocks in the DPB 218. The motion estimation unit 222 and the motion compensation unit 224 can retrieve a reference image formed by the reconstructed (and possibly filtered) blocks from the DPB 218 to perform inter-frame prediction of blocks in subsequently encoded images. Additionally, the intra-frame prediction unit 226 can use the reconstructed blocks in the DPB 218 of the current image to intra-frame predict other blocks in the current image.

[0130] Typically, 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), probabilistic interval partitioned entropy (PIPE) decoding, exponential-Golomb decoding, or other types of entropy coding operations on the data. In some examples, the entropy coding unit 220 can operate in a bypass mode, where syntax elements are not entropy encoded.

[0131] The video encoder 200 can output a bitstream that includes entropy-encoded syntax elements required to reconstruct slices or blocks of images. Specifically, the entropy coding unit 220 can output a bitstream.

[0132] The operations described above pertain to block descriptions. This description should be understood as the operation of the luma decoding block and / or chroma decoding block. 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.

[0133] In some examples, the operations performed for the luma decoded block do not need to be repeated for the chroma decoded block. As an example, the operations for identifying the motion vector (MV) and reference image of the luma decoded block do not need to be repeated for identifying the MV and reference image of the chroma block. Instead, the MV of the luma decoded block can be scaled to determine the MV of the chroma block, and the reference image can be the same. As another example, the intra-frame prediction processing can be the same for both the luma and chroma decoded blocks.

[0134] Figure 7 This is a block diagram illustrating an example video decoder 300 that can perform the techniques disclosed herein. Figure 7 This disclosure is provided for illustrative purposes and does not limit the techniques extensively illustrated and described herein. For illustrative purposes, this disclosure describes the video decoder 300 according to the techniques of VVC and HEVC (ITU-T H.265). However, the techniques of this disclosure can be implemented by video decoding devices configured for other video decoding standards.

[0135] exist Figure 7 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 decoded picture buffer (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 processing circuits. For example, 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.

[0136] 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 to 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.

[0137] CPB memory 320 can store video data to be decoded by components of video decoder 300, such as encoded video bitstreams. For example, video data stored in CPB memory 320 can be obtained, for example, from computer-readable medium 110 (…). Figure 1The CPB memory 320 may include a CPB that stores encoded video data (e.g., syntax elements) from the encoded video bitstream. Additionally, 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 decoded pictures that the video decoder 300 may output and / or use as reference video data when decoding subsequent data or pictures from the encoded video bitstream. The CPB memory 320 and DPB 314 may be formed of 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 CPB memory 320 and 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 along with other components of the video decoder 300, or off-chip relative to these components.

[0138] Additionally or alternatively, in some examples, the video decoder 300 can be drawn from the memory 120 ( Figure 1 The decoded video data is retrieved. That is, memory 120 can store data using CPB memory 320 as described above. Similarly, when some or all of the functions of video decoder 300 are implemented in software executed by the processing circuitry of video decoder 300, memory 120 can store instructions to be executed by video decoder 300.

[0139] Figure 7 The various units shown are illustrated to aid in understanding 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 6 Fixed-function circuits refer to circuits that provide a specific function and have pre-defined executable operations. Programmable circuits refer to circuits that can be programmed to perform various tasks and provide flexible functionality in the executable operations. For example, programmable circuits can execute software or firmware, causing 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 usually immutable. In some examples, one or more units may be different circuit blocks (fixed-function or programmable), and in some examples, one or more units may be integrated circuits.

[0140] The video decoder 300 may include an ALU, an 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 the 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.

[0141] Entropy decoding unit 302 can receive encoded video data from the CPB and perform entropy decoding on the video data to reproduce the 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 the syntax elements extracted from the bitstream.

[0142] Typically, the video decoder 300 reconstructs the image on a block-by-block basis. The video decoder 300 can perform the reconstruction operation separately for each block (where the block currently being reconstructed (i.e., decoded) can be referred to as the "current block").

[0143] 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 (multiple) transform mode indicators. 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. For example, inverse quantization unit 306 can perform a bit-by-bit left shift operation to inverse quantize the quantized transform coefficients. Inverse quantization unit 306 can thus form a transform coefficient block including the transform coefficients.

[0144] After the inverse quantization unit 306 forms the transform coefficient block, the inverse transform processing unit 308 can 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 can apply inverse DCT, inverse integer transform, inverse Karl Henle-Loughlin transform (KLT), inverse rotation transform, inverse direction transform, or other inverse transforms to the transform coefficient block.

[0145] Furthermore, prediction processing unit 304 generates prediction blocks based on prediction information syntax elements entropy decoded by entropy decoding unit 302. For example, if the prediction information syntax elements indicate that the current block is inter-frame predicted, motion compensation unit 316 can generate prediction blocks. In this case, the prediction information syntax elements may indicate a reference image in DPB 314 from which a reference block is retrieved, and a motion vector identifying the position of the reference block in the reference image relative to the position of the current block in the current image. Motion compensation unit 316 can generally be configured in a manner substantially similar to that of motion compensation unit 224 ( Figure 6 The method described is used to perform inter-frame prediction processing.

[0146] As another example, if the prediction information syntax element indicates that the current block is intra-predictive, then intra-predictive unit 318 can generate a prediction block according to the intra-predictive mode indicated by the prediction information syntax element. Similarly, intra-predictive unit 318 can generally be configured in a manner substantially similar to that of intra-predictive unit 226 ( Figure 6 Intra-prediction processing is performed in the manner described. Intra-prediction unit 318 can retrieve data of neighboring samples of the current block from DPB 314.

[0147] Reconstruction unit 310 can reconstruct the current block using the prediction block and the residual 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.

[0148] Filter unit 312 can perform one or more filtering operations on the reconstructed block. For example, filter unit 312 can perform a deblocking operation to reduce block artifacts along the edges of the reconstructed block. The operation of filter unit 312 need not be performed in all examples. Filter unit 312 can be configured to apply techniques of this disclosure, such as resizing the color component to a minimum color component size, and then applying a convolutional neural network (CNN) filter 322 to the resized color component.

[0149] Specifically, filter unit 312 can apply one or more CNN filters 322 to reduce the size of the luminance component to the size of a smaller chrominance component in the chrominance components, instead of upsampling the chrominance component to the size of the corresponding luminance component. That is, filter unit 312 can receive a luminance component with a first size and a chrominance component with a second size smaller than the first size. Filter unit 312 can apply a convolutional NN filter of the CNN filters 322 to the luminance component to downsample the luminance component to the second size. Filter unit 312 can filter the chrominance component to form a filtered chrominance component. Then, filter unit 312 can cascade the downsampled luminance component and the filtered chrominance component to form a cascaded color component, and then filter the cascaded color component to form a filtered cascaded component, which includes the filtered downsampled first color component. Furthermore, filter unit 312 can upsample the filtered luminance component of the second size back to the first size before storing the luminance component in DPB 218.

[0150] Specifically, the first and second color components may initially be stored in separate arrays or matrices. For cascaded color components, filter unit 312 can form a single array or matrix with a width or height twice that of a single color component. Filter unit 312 can then store samples of the first color component in a first region of the newly formed array or matrix and samples of the second color component in an adjacent second region of the newly formed array or matrix. When using three color components (e.g., luminance, blue hue chroma, and red hue chroma), filter unit 312 can form a single array or matrix with a width or height three times that of a single color component and store samples of each of the three color components in the corresponding region of the newly formed array or matrix.

[0151] The video decoder 300 can store reconstructed blocks in the DPB 314. For example, in an example where the filter unit 312 is not operated, the reconstruction unit 310 can store the reconstructed blocks in the DPB 314. In an example where the filter unit 312 is operated, the filter unit 312 can store the filtered reconstructed blocks in the DPB 314. As described above, the DPB 314 can provide reference information to the prediction processing unit 304, such as samples of the current image for intra-frame prediction and previously decoded images for subsequent motion compensation. Furthermore, the video decoder 300 can output the decoded image from the DPB for subsequent display on a display device (e.g., ...). Figure 1 The display device 118) presents the image.

[0152] Figure 8 This is a block diagram illustrating an example of a joint component convolutional neural network (CNN) filter unit 330 according to the technology of this disclosure. Figure 8 The CNN filter can correspond to the one executed by filter unit 216. Figure 6 CNN filter 232 or by Figure 7 The filter unit 312 executes the CNN filter 322.

[0153] In this example, filter unit 330 (e.g., filter unit 216 or 312) includes CNN filters 332, 334, 340, 342, and 344, cascade unit 336, hidden layers 338A-338N (hidden layer 338), and channel adders 346, 348, and 350. CNN filter 332 can downsample the luminance component with a convolutional neural network (CNN) layer with a stride of 2 in both the horizontal and vertical directions. That is, CNN filter 332 can apply a 3×3×M convolutional neural network layer filter with a stride of (2,2) to the luminance component, and then apply the PReLU activation function to the filtered sample of the luminance component.

[0154] Then, CNN filter 334 filters the chroma components (i.e., Cb and Cr) and applies the PReLU activation function to the filtered samples of the chroma components. Cascade unit 336 then cascades the output of the downsampled luminance component with the output of CNN filter 334 to form cascaded color components. Hidden layer 338 can then use the cascaded color components as input. In the final layer, CNN filter 340 upsamples the output from hidden layer 338 and feeds the upsampled data back to CNN filter 340 to generate the luminance component output. Similarly, CNN filters 342 and 344 filter the corresponding chroma components of the output from hidden layer 338. Individual channel adders 346, 348, and 350 separate the cascaded color components into their respective luminance and chroma components.

[0155] Figure 8 The blocks of hidden layer 338 can be combinations of filters or functions. For example, each block in hidden layer 338 can include a convolutional neural network filter plus an activation function or such as... Figure 9 The residual processing unit shown.

[0156] Figure 9 This is a conceptual diagram illustrating an example of a residual processing unit 360. The residual processing unit 360 can correspond to... Figure 8 The hidden layer 338 comprises one or more blocks. In this example, the residual processing unit 360 includes a first 3×3×K×K convolutional layer 362, a parametric rectified linear unit (PReLU) layer 364, a second 3×3×K×K convolutional layer 366, and a summing unit 368. The first 3×3×K×K convolutional layer 362 applies a first CNN filter to the received samples, the PReLU layer 364 applies a PReLU activation function to the filtered samples, and the second 3×3×K×K convolutional layer 366 applies a second CNN filter to the filtered samples from the PReLU layer 364. Finally, the summing unit 368 combines the filtered samples with the received input samples.

[0157] Figure 10 This is a block diagram of another example of a joint component CNN filter unit 330' shown according to the technology of this disclosure. Figure 10 The CNN filter unit 330' can correspond to the one made by Figure 6 The CNN filter 232 is executed by the filter unit 216 or by the filter unit 216. Figure 7 The filter unit 312 executes the CNN filter 322.

[0158] Figure 10 Example CNN filter unit 330' and Figure 8The CNN filter unit 330 is essentially similar. However, in this example, the CNN filter unit 330' includes a block combination unit 370 instead of... Figure 8 The CNN filter unit 330 has a CNN filter 340. The block combination unit 370 can combine four N×N blocks to form a 2N×2N luminance block output from the hidden layer set. These four N×N blocks are the output of the last convolutional NN layer, where the last convolutional NN layer can also be a "ConvTranspose2d / De-convolution" operation with a stride of 2 in both the horizontal and vertical directions.

[0159] Figure 11 This is a flowchart illustrating an example method for encoding the current block according to the technology of this disclosure. The current block may include the current CU. Although regarding video encoder 200 ( Figure 1 and Figure 3 This is described in detail, but it should be understood that other devices can be configured to perform similar actions. Figure 11 A similar approach.

[0160] In this example, the video encoder 200 initially predicts the current block (380). 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 (382). To compute the residual block, the video encoder 200 may compute the difference between the original, uncoded block and the prediction block for the current block. The video encoder 200 may then transform and quantize the coefficients of the residual block (384). Next, the video encoder 200 may scan the quantized transform coefficients of the residual block (386). During or after the scan, the video encoder 200 may entropy encode the coefficients (388). For example, the video encoder 200 may encode the coefficients using CAVLC or CABAC. The video encoder 200 may then output the entropy-encoded data for the block (390).

[0161] The video encoder 200 can also decode the current block after encoding it, using the decoded version of the current block as reference data for subsequent decoding (e.g., in inter-frame prediction or intra-frame prediction mode). Therefore, the video encoder 200 can perform inverse quantization and inverse transform on the coefficients to reproduce the residual block (392). The video encoder 200 can combine the residual block with the prediction block to form a decoded block (394).

[0162] According to the technology disclosed herein, video encoder 200 can filter and refine decoded blocks (396). For example, video encoder 200 can downsample the luma component (e.g., using a convolutional neural network filter). Video encoder 200 can also filter the chroma component and then concatenate the downsampled luma component with the filtered chroma component. Video encoder 200 can also filter the concatenated color components to form a filtered concatenated component, wherein the filtered concatenated component includes a filtered downsampled first color component. Video encoder 200 can also filter the concatenated color components and then upsample the filtered luma component to its original size. Video encoder 200 can then store the filtered decoded block in DPB 218 (398).

[0163] In this way, Figure 11 The method illustrates an example of filtering decoded video data, comprising: applying a downsampled convolutional neural network layer to a first color component of a block of video data, the first color component having a first size, wherein applying the downsampled convolutional neural network layer to the first color component generates a downsampled first color component having a second size smaller than the first size; filtering the second color component having the second size to form a filtered second color component; concatenating the downsampled first color component and the filtered second color component to form a concatenated color component; and filtering the concatenated color component.

[0164] Figure 12 This is a flowchart illustrating an example method for decoding the current block according to the technology of this disclosure. The current block may include the current CU. Although regarding the video decoder 300 ( Figure 1 As described in Figure 4), however, it should be understood that other devices can be configured to perform the same actions. Figure 12 A similar approach.

[0165] The video decoder 300 can receive entropy-coded data of the current block, such as entropy-coded prediction information and entropy-coded data of the coefficients of the residual block corresponding to the current block (400). The video decoder 300 can entropy decode the entropy-coded data to determine the prediction information of the current block and reproduce the coefficients of the residual block (402). The video decoder 300 can predict the current block (404), for example, using an intra-frame prediction or inter-frame prediction mode indicated by the prediction information of the current block to compute the prediction block of the current block. The video decoder 300 can then inversely scan the reproduced coefficients (406) to create a quantized transform coefficient block. The video decoder 300 can then inversely quantize and inverse transform the quantized transform coefficients to produce the residual block (408). The video decoder 300 can finally decode the current block by combining the prediction block and the residual block (410).

[0166] The video decoder 300 can also filter the decoded video data (412), for example, using a convolutional neural network filter as discussed above, according to the techniques of this disclosure. For example, the video decoder 300 can downsample the luminance component (e.g., using a convolutional neural network filter). The video decoder 300 can also filter the chrominance component and then concatenate the downsampled luminance component with the filtered chrominance component. The video decoder 300 can also filter the concatenated chrominance component. The video decoder 300 can also filter the concatenated chrominance component and then upsample the filtered luminance component to its original size. The video decoder 300 can then store the filtered decoded block in DPB218 (414).

[0167] In this way, Figure 12 The method illustrates an example of filtering decoded video data, comprising: applying a downsampled convolutional neural network layer to a first color component of a block of video data, the first color component having a first size, wherein applying the downsampled convolutional neural network layer to the first color component generates a downsampled first color component having a second size smaller than the first size; filtering the second color component having the second size to form a filtered second color component; concatenating the downsampled first color component and the filtered second color component to form a concatenated color component; and filtering the concatenated color component.

[0168] Figure 13 This is a flowchart illustrating an example method for filtering decoded video data according to the technology of this disclosure. A video encoder 200, a video decoder 300, or a video decoder 130 can perform this operation. Figure 13 The method. For illustrative purposes, regarding video decoder 300... Figure 13 Explanation of the method.

[0169] Initially, the video decoder 300 can receive and decode video data. The decoded video data may include a luminance component and two chrominance components (e.g., Cb and Cr data). In this example, it is assumed that each of these components has a different size, i.e., the luminance component has a first size, the first chrominance component has a second size, and the second chrominance component has a third size, where the second size is smaller than the first size, and the third size is smaller than the second size. For example, in this example, the video data may have a 4:2:0 format.

[0170] The video decoder 300 may apply a downsampled CNN layer to the luminance component (420), which produces a downsampled luminance component with a third size. The video decoder 300 may also apply a downsampled CNN layer (or a different downsampled CNN layer) to the first chrominance component (422), which produces a first downsampled chrominance component with a third size. The video decoder 300 may then filter the first downsampled chrominance component and the second chrominance component (424) (e.g., using a CNN filter).

[0171] Then, the video decoder 300 can concatenate the luminance component and the chrominance component (426), which produces a concatenated color component. The video decoder 300 can then filter the concatenated color component (428), (e.g., using a CNN filter). Figure 8 and Figure 10 Hidden layer 338).

[0172] The video decoder 300 can then upsample the luminance component of the filtered cascaded color components to a first size (430) and upsample the first chrominance component of the filtered cascaded color components to a second size (432). The video decoder 300 can then store the components in the DPB 314 and output these components (434).

[0173] In this way, Figure 13 The method illustrates an example of filtering decoded video data, comprising: applying a downsampled convolutional neural network layer to a first color component of a block of video data, the first color component having a first size, wherein applying the downsampled convolutional neural network layer to the first color component generates a downsampled first color component having a second size smaller than the first size; filtering the second color component having the second size to form a filtered second color component; concatenating the downsampled first color component and the filtered second color component to form a concatenated color component; and filtering the concatenated color component.

[0174] Figure 14 This is a flowchart illustrating an example method for filtering decoded video data according to the technology of this disclosure. A video encoder 200, a video decoder 300, or a video decoder 130 can perform this operation. Figure 14 The method. For illustrative purposes, regarding video decoder 300... Figure 14 Explanation of the method.

[0175] Initially, the video decoder 300 can receive and decode video data. The decoded video data may include a luminance component and two chrominance components (e.g., Cb and Cr data). In this example, it is assumed that the luminance component has a first size, and the first and second chrominance components have a second size, wherein the second size is smaller than the first size. For example, in this example, the video data may be in 4:2:0 or 4:2:2 format.

[0176] The video decoder 300 can apply a downsampled CNN layer to the luminance component (440), which produces a downsampled luminance component with a second size. The video decoder 300 can then filter the first chrominance component and the second chrominance component (442) (e.g., using a CNN filter).

[0177] The video decoder 300 can then concatenate the luminance and chrominance components (444), which produces a concatenated color component. The video decoder 300 can then filter the concatenated color component (446), for example, using a CNN filter (such as...). Figure 8 and Figure 10 Hidden layer 338).

[0178] The video decoder 300 can then upsample the luminance component of the filtered, cascaded color components to a first size (448). The video decoder 300 can then store the component in a DPB 314 and output the component (450).

[0179] In this way, Figure 14 The method illustrates an example of filtering decoded video data, comprising: applying a downsampled convolutional neural network layer to a first color component of a block of video data, the first color component having a first size, wherein applying the downsampled convolutional neural network layer to the first color component generates a downsampled first color component having a second size smaller than the first size; filtering the second color component having the second size to form a filtered second color component; concatenating the downsampled first color component and the filtered second color component to form a concatenated color component; and filtering the concatenated color component.

[0180] Some examples of the technology disclosed herein are summarized in the following clauses:

[0181] Clause 1: A method for filtering decoded video data, the method comprising: applying a downsampled convolutional neural network layer to a first color component of a block of video data, the first color component of the block having a first size, wherein applying the downsampled convolutional neural network layer to the first color component produces a downsampled first color component having a second size; filtering the downsampled first color component using one or more convolutional neural network layer filters; and filtering a second color component of the video data block using the one or more convolutional neural network layer filters.

[0182] Clause 2: The method according to Clause 1 further includes upsampling the filtered, downsampled first color component to the first size.

[0183] Clause 3: The method according to Clause 1 further includes combining two or more filtered downsampled blocks (including filtered downsampled blocks) of the first color component to generate an upsampled first color component having the first size.

[0184] Clause 4: According to the method described in Clause 3, wherein the first size comprises 2N×2N, and wherein the two or more filtered downsampled blocks of the first color component comprise four N×N filtered downsampled blocks of the first color component.

[0185] Clause 5: The method according to any one of Clauses 1-4, wherein the downsampling convolutional neural network layer comprises a 3×3×M convolutional neural network layer filter with a stride of 2.

[0186] Clause 6: The method according to any one of Clauses 1-5, wherein the one or more convolutional neural network layer filters include residual processing units.

[0187] Clause 7: The method according to Clause 6, wherein the residual processing unit comprises a first 3×3×K×K convolutional layer, a PReLU layer, and a second 3×3×K×K convolutional layer.

[0188] Clause 8: The method according to any one of Clauses 1-7 further includes cascading the downsampled first color component and the second color component.

[0189] Clause 9: The method according to any one of Clauses 1-8 further includes filtering the third color component of the video data block using the one or more convolutional neural network layer filters.

[0190] Clause 10: The method according to Clause 9, wherein the second dimension includes the smaller of the dimension of the second color component or the dimension of the third color component.

[0191] Clause 11: The method according to any one of Clauses 1-10, wherein the first color component includes a luminance component, and the second color component includes one of a blue hue chromaticity component or a red hue chromaticity component.

[0192] Clause 12: An apparatus for decoding video data, the apparatus comprising one or more components for performing the method described in any one of Clauses 1-11.

[0193] Clause 13: The device pursuant to Clause 12, wherein the one or more components include one or more processors implemented in a circuit.

[0194] Clause 14: The device according to Clause 12 further includes a display configured to display the decoded video data.

[0195] Clause 15: The device as described in Clause 12, wherein the device includes one or more of a camera, computer, mobile device, broadcast receiving device or set-top box.

[0196] Clause 16: The device described in Clause 12 further includes a memory configured to store video data.

[0197] Clause 17: A computer-readable storage medium having instructions stored thereon that, when executed, cause a processor to perform any one of Clauses 1-11.

[0198] Clause 18: A method for filtering decoded video data, the method comprising: applying a downsampled convolutional neural network layer to a first color component of a block of video data, the first color component of the block having a first size, wherein applying the downsampled convolutional neural network layer to the first color component generates a downsampled first color component having a second size smaller than the first size; filtering the second color component having the second size to form a filtered second color component; concatenating the downsampled first color component and the filtered second color component to form a concatenated color component; and filtering the concatenated color component.

[0199] Clause 19: The method according to Clause 18 further includes upsampling the filtered, downsampled first color component to the first size.

[0200] Clause 20: The method according to Clause 18 further includes combining two or more filtered downsampled blocks (including the filtered downsampled blocks) of the first color component to generate an upsampled first color component having the first size.

[0201] Clause 21: The method according to Clause 20, wherein the first size comprises 2N×2N, and wherein the two or more filtered downsampled blocks of the first color component comprises four N×N filtered downsampled blocks of the first color component.

[0202] Clause 22: The method according to Clause 18, wherein the downsampling convolutional neural network layer comprises a 3×3×M convolutional neural network layer filter with a stride of 2.

[0203] Clause 23: The method according to Clause 18, wherein the one or more convolutional neural network layer filters include residual processing units.

[0204] Clause 24: The method according to Clause 23, wherein the residual processing unit comprises a first 3×3×K×K convolutional layer, a PReLU layer, and a second 3×3×K×K convolutional layer.

[0205] Clause 25: The method according to Clause 18 further includes filtering the third color component of the video data block using the one or more convolutional neural network layer filters.

[0206] Clause 26: The method according to Clause 25, wherein the second dimension includes the smaller of the dimension of the second color component or the dimension of the third color component.

[0207] Clause 27: The method according to Clause 18, wherein the first color component includes a luminance component, and the second color component includes either a blue hue chromaticity component or a red hue chromaticity component.

[0208] Clause 28: The method according to Clause 18, wherein filtering the second color component includes filtering the second color component using a convolutional neural network filter.

[0209] Clause 29: The method according to Clause 18, wherein filtering the cascaded color components includes filtering the cascaded color components using a convolutional neural network filter.

[0210] Clause 30: An apparatus for decoding video data, the apparatus comprising: a memory configured to store the video data; and one or more processors implemented in a circuit and configured to: apply a downsampled convolutional neural network layer to a first color component of a block of video data, the first color component of the block having a first size, wherein applying the downsampled convolutional neural network layer to the first color component generates a downsampled first color component having a second size smaller than the first size; filter the second color component having the second size to form a filtered second color component; concatenate the downsampled first color component and the filtered second color component to form a concatenated color component; and filter the concatenated color component.

[0211] Clause 31: The device according to Clause 30, wherein the one or more processors are further configured to upsample the filtered, downsampled first color component to the first size.

[0212] Clause 32: The device according to Clause 30, wherein the one or more processors are further configured to combine two or more filtered downsampled blocks (including the filtered downsampled blocks) of the first color component to generate an upsampled first color component having the first size.

[0213] Clause 33: The device according to Clause 32, wherein the first size comprises 2N×2N, and wherein the two or more filtered downsampled blocks of the first color component comprise four N×N filtered downsampled blocks of the first color component.

[0214] Clause 34: The device according to Clause 30, wherein the downsampling convolutional neural network layer comprises a 3×3×M convolutional neural network layer filter with a stride of 2.

[0215] Clause 35: The device according to Clause 30, wherein the one or more convolutional neural network layer filters include a residual processing unit.

[0216] Clause 36: The apparatus according to Clause 35, wherein the residual processing unit comprises a first 3×3×K×K convolutional layer, a PReLU layer, and a second 3×3×K×K convolutional layer.

[0217] Clause 37: The device according to Clause 30, wherein the one or more processors are further configured to filter the third color component of the video data block using the one or more convolutional neural network layer filters.

[0218] Clause 38: The device pursuant to Clause 37, wherein the second dimension includes the smaller of the dimension of the second color component or the dimension of the third color component.

[0219] Clause 39: The device according to Clause 30, wherein the first color component includes a luminance component, and the second color component includes either a blue hue chromaticity component or a red hue chromaticity component.

[0220] Clause 40: The device according to Clause 30, wherein the one or more processors are configured to filter the second color component using a convolutional neural network filter.

[0221] Clause 41: The device according to Clause 30, wherein the one or more processors are configured to filter the cascaded color components using a convolutional neural network filter.

[0222] Clause 42: The device according to Clause 30 further includes a display configured to display video data corresponding to the cascaded color components.

[0223] Clause 43: The device described in Clause 30, wherein the device includes one or more of a camera, computer, mobile device, broadcast receiver device, or set-top box.

[0224] Clause 44: A computer-readable storage medium having instructions stored thereon, which, when executed, cause a processor to: apply a downsampled convolutional neural network layer to a first color component of a block of video data, the first color component of the block having a first size, wherein applying the downsampled convolutional neural network layer to the first color component generates a downsampled first color component having a second size smaller than the first size; filter the second color component having the second size to form a filtered second color component; concatenate the downsampled first color component and the filtered second color component to form a concatenated color component; and filter the concatenated color component.

[0225] Clause 45: The computer-readable storage medium according to Clause 44 further includes instructions that cause the processor to upsample the filtered, downsampled first color component to the first size.

[0226] Clause 46: The computer-readable storage medium according to Clause 44 further includes instructions that cause the processor to combine two or more filtered downsampled blocks (including filtered downsampled blocks) of the first color component to generate an upsampled first color component having the first size.

[0227] Clause 47: The computer-readable storage medium according to Clause 46, wherein the first size comprises 2N×2N, and wherein the two or more filtered downsampled blocks of the first color component comprise four N×N filtered downsampled blocks of the first color component.

[0228] Clause 48: The computer-readable storage medium according to Clause 44, wherein the downsampled convolutional neural network layer comprises a 3×3×M convolutional neural network layer filter with a stride of 2.

[0229] Clause 49: The computer-readable storage medium according to Clause 44, wherein the one or more convolutional neural network layer filters include a first 3×3×K×K convolutional layer, a PReLU layer, and a second 3×3×K×K convolutional layer.

[0230] Clause 50: A computer-readable storage medium pursuant to Clause 44, wherein the one or more processors are further configured to filter the third color component of the video data block using the one or more convolutional neural network layer filters.

[0231] Clause 51: A computer-readable storage medium pursuant to Clause 50, wherein the second size includes the smaller of the size of the second color component or the size of the third color component.

[0232] Clause 52: The computer-readable storage medium pursuant to Clause 44, wherein the first color component comprises a luminance component, and the second color component comprises either a blue hue chromaticity component or a red hue chromaticity component.

[0233] Clause 53: A computer-readable storage medium according to Clause 44, wherein the instructions for causing the processor to filter the second color component include instructions for causing the processor to filter the second color component using a convolutional neural network filter.

[0234] Clause 54: The computer-readable storage medium according to Clause 44, wherein the instructions for causing the processor to filter the second color component include instructions for causing the processor to filter the cascaded color components using a convolutional neural network filter.

[0235] Clause 55: An apparatus for filtering decoded video data, the apparatus comprising: means for applying a downsampled convolutional neural network layer to a first color component of a block of video data, the first color component of the block having a first size, wherein applying the downsampled convolutional neural network layer to the first color component generates a downsampled first color component having a second size smaller than the first size; means for filtering the second color component having the second size to form a filtered second color component; means for cascading the downsampled first color component and the filtered second color component to form a cascaded color component; and means for filtering the cascaded color component.

[0236] Clause 56: The apparatus according to Clause 55, wherein the components for filtering the second color component include components for filtering the second color component using a convolutional neural network filter.

[0237] Clause 57: The apparatus according to Clause 55, wherein the components for filtering the cascaded color components include components for filtering the cascaded color components using a convolutional neural network filter.

[0238] Clause 58: A method for filtering decoded video data, the method comprising: applying a downsampled convolutional neural network layer to a first color component of a block of video data, the first color component of the block having a first size, wherein applying the downsampled convolutional neural network layer to the first color component generates a downsampled first color component having a second size smaller than the first size; filtering the second color component having the second size to form a filtered second color component; concatenating the downsampled first color component and the filtered second color component to form a concatenated color component; and filtering the concatenated color component.

[0239] Clause 59: The method according to Clause 58 further includes upsampling the filtered, downsampled first color component to the first size.

[0240] Clause 60: The method according to Clause 58 further includes combining two or more filtered downsampled blocks (including the filtered downsampled blocks) of the first color component to generate an upsampled first color component having the first size.

[0241] Clause 61: The method according to Clause 60, wherein the first size comprises 2N×2N, and wherein the two or more filtered downsampled blocks of the first color component comprise four N×N filtered downsampled blocks of the first color component.

[0242] Clause 62: The method according to any one of Clauses 58-61, wherein the downsampling convolutional neural network layer comprises a 3×3×M convolutional neural network layer filter with a stride of 2.

[0243] Clause 63: The method according to any one of Clauses 58-62, wherein the one or more convolutional neural network layer filters include residual processing units.

[0244] Clause 64: The method according to Clause 63, wherein the residual processing unit comprises a first 3×3×K×K convolutional layer, a PReLU layer, and a second 3×3×K×K convolutional layer.

[0245] Clause 65: The method according to any one of Clauses 58-64 further includes filtering the third color component of the video data block using the one or more convolutional neural network layer filters.

[0246] Clause 66: The method according to Clause 65, wherein the second dimension includes the smaller of the dimension of the second color component or the dimension of the third color component.

[0247] Clause 67: The method according to any one of Clauses 58-66, wherein the first color component includes a luminance component, and wherein the second color component includes one of a blue hue chromaticity component or a red hue chromaticity component.

[0248] Clause 68: The method according to any one of Clauses 58-67, wherein filtering the second color component includes filtering the second color component using a convolutional neural network filter.

[0249] Clause 69: The method according to any one of Clauses 58-68, wherein filtering the cascaded color components includes filtering the cascaded color components using a convolutional neural network filter.

[0250] Clause 70: An apparatus for decoding video data, the apparatus comprising: a memory configured to store the video data; and one or more processors implemented in a circuit and configured to: apply a downsampled convolutional neural network layer to a first color component of a block of video data, the first color component of the block having a first size, wherein applying the downsampled convolutional neural network layer to the first color component generates a downsampled first color component having a second size smaller than the first size; filter the second color component having the second size to form a filtered second color component; concatenate the downsampled first color component and the filtered second color component to form a concatenated color component; and filter the concatenated color component.

[0251] Clause 71: The device according to Clause 70, wherein the one or more processors are further configured to upsample the filtered, downsampled first color component to the first size.

[0252] Clause 72: The device according to Clause 70, wherein the one or more processors are further configured to combine two or more filtered downsampled blocks (including the filtered downsampled blocks) of the first color component to generate an upsampled first color component having the first size.

[0253] Clause 73: The device according to Clause 72, wherein the first dimension comprises 2N×2N, and wherein the two or more filtered downsampled blocks of the first color component comprise four N×N filtered downsampled blocks of the first color component.

[0254] Clause 74: The device according to any one of Clauses 70-73, wherein the downsampling convolutional neural network layer comprises a 3×3×M convolutional neural network layer filter with a stride of 2.

[0255] Clause 75: The device according to any one of Clauses 70-74, wherein the one or more convolutional neural network layer filters include a residual processing unit.

[0256] Clause 76: The apparatus according to Clause 75, wherein the residual processing unit comprises a first 3×3×K×K convolutional layer, a PReLU layer, and a second 3×3×K×K convolutional layer.

[0257] Clause 77: The device according to any one of Clauses 70-76, wherein the one or more processors are further configured to filter the third color component of the video data block using one or more convolutional neural network layer filters.

[0258] Clause 78: The device pursuant to Clause 77, wherein the second dimension includes the smaller of the dimension of the second color component or the dimension of the third color component.

[0259] Clause 79: The device according to any one of Clauses 70-78, wherein the first color component includes a luminance component, and wherein the second color component includes either a blue hue chromaticity component or a red hue chromaticity component.

[0260] Clause 80: The device according to any one of Clauses 70-79, wherein the one or more processors are configured to filter the second color component using a convolutional neural network filter.

[0261] Clause 81: A device according to any one of Clauses 70-80, wherein the one or more processors are configured to filter the cascaded color components using the convolutional neural network filter.

[0262] Clause 82: The device according to any one of Clauses 70-81 further includes a display configured to display video data corresponding to the cascaded color components.

[0263] Clause 83: The device pursuant to any one of Clauses 70-82, wherein the device includes one or more of a camera, computer, mobile device, broadcast receiver device, or set-top box.

[0264] Clause 84: A computer-readable storage medium having instructions stored thereon, which, when executed, cause a processor to: apply a downsampled convolutional neural network layer to a first color component of a block of video data, the first color component of the block having a first size, wherein applying the downsampled convolutional neural network layer to the first color component generates a downsampled first color component having a second size smaller than the first size; filter the second color component having the second size to form a filtered second color component; concatenate the downsampled first color component and the filtered second color component to form a concatenated color component; and filter the concatenated color component.

[0265] Clause 85: The computer-readable storage medium according to Clause 84 further includes instructions to cause the processor to upsample the filtered, downsampled first color component to the first size.

[0266] Clause 86: The computer-readable storage medium according to Clause 84 further includes instructions that cause the processor to combine two or more filtered downsampled blocks (including filtered downsampled blocks) of the first color component to generate an upsampled first color component having the first size.

[0267] Clause 87: A computer-readable storage medium pursuant to Clause 86, wherein the first size comprises 2N × 2N, and wherein the two or more filtered downsampled blocks of the first color component comprise four N × N filtered downsampled blocks of the first color component.

[0268] Clause 88: A computer-readable storage medium according to any one of Clauses 84-87, wherein the downsampled convolutional neural network layer comprises a 3×3×M convolutional neural network layer filter with a stride of 2.

[0269] Clause 89: A computer-readable storage medium according to any one of Clauses 84-88, wherein the one or more convolutional neural network layer filters comprise a first 3×3×K×K convolutional layer, a PReLU layer, and a second 3×3×K×K convolutional layer.

[0270] Clause 90: A computer-readable storage medium according to any one of Clauses 84-89, wherein the one or more processors are further configured to filter the third color component of the video data block using the one or more convolutional neural network layer filters.

[0271] Clause 91: A computer-readable storage medium pursuant to Clause 90, wherein the second size includes the smaller of the size of the second color component or the size of the third color component.

[0272] Clause 92: A computer-readable storage medium pursuant to any one of Clauses 84-91, wherein the first color component comprises a luminance component, and wherein the second color component comprises one of a blue hue chromaticity component or a red hue chromaticity component.

[0273] Clause 93: A computer-readable storage medium according to any one of Clauses 84-92, wherein the instructions for causing the processor to filter the second color component include instructions for causing the processor to filter the second color component using a convolutional neural network filter.

[0274] Clause 94: A computer-readable storage medium according to any one of Clauses 84-93, wherein the instructions for causing the processor to filter the second color component include instructions for causing the processor to filter the cascaded color components using a convolutional neural network filter.

[0275] Clause 95: An apparatus for filtering decoded video data, the apparatus comprising: means for applying a downsampled convolutional neural network layer to a first color component of a block of video data, the first color component of the block having a first size, wherein applying the downsampled convolutional neural network layer to the first color component generates a downsampled first color component having a second size smaller than the first size; means for filtering the second color component having the second size to form a filtered second color component; means for concatenating the downsampled first color component and the filtered second color component to form a concatenated color component; and means for filtering the concatenated color component.

[0276] Clause 96: The apparatus according to Clause 95, wherein the components for filtering the second color component include components for filtering the second color component using a convolutional neural network filter.

[0277] Clause 97: The apparatus according to any one of Clauses 95 and 96, wherein the components for filtering the cascaded color components include components for filtering the cascaded color components using a convolutional neural network filter.

[0278] It should be recognized that, depending on the examples, certain actions or events of any technique described herein may be performed in a different order, or may be added, combined, or omitted entirely (e.g., not all described actions or events are necessary for the practice of the technique). Furthermore, in some examples, actions or events may be performed concurrently, for example, through multithreading, interrupt handling, or multiple processors, rather than sequentially.

[0279] In one or more examples, the described functionality may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, these functions may be stored or transmitted as one or more instructions or code on or through a computer-readable medium and executed by a hardware-based processing unit. A computer-readable medium may include a computer-readable storage medium corresponding to a tangible medium such as a data storage medium, or a communication medium that includes any medium that facilitates (e.g., according to a communication protocol) the transfer of a computer program from one place to another. 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 accessible by one or more computers or one or more processors to retrieve instructions, code, and / or data structures to implement the techniques described in this disclosure. Computer program products may include computer-readable media.

[0280] By way of example and not limitation, such computer-readable storage media may include RAM, ROM, EEPROM, CD-ROM or other optical disc storage, disk storage, or other magnetic storage devices, flash memory, or any other medium that can be used to store desired program code in the form of instructions or data structures and that can be accessed by a computer. Furthermore, any connection is properly 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 (e.g., infrared, radio, and microwave), then coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technology (e.g., 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 rather refer to non-transient tangible storage media. Disks and optical discs as used herein include compact optical discs (CDs), laser discs, optical discs, digital versatile discs (DVDs), floppy disks, and Blu-ray discs, where disks typically reproduce data magnetically, while optical discs reproduce data optically using lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0281] Instructions can be executed by one or more processors, such as one or more digital signal processors (DSPs), general-purpose microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (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 in 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.

[0282] The techniques disclosed herein can be implemented in a 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 they do not necessarily need to be implemented through different hardware units. Rather, as described above, various units can be combined in a codec hardware unit, or provided by a collection of interoperable hardware units including one or more processors as described above, combined with suitable software and / or firmware.

[0283] Various examples have been described. These and other examples are all within the scope of the appended claims.

Claims

1. A method for filtering decoded video data, the method comprising: A downsampled convolutional neural network layer is applied to a first color component of a video data block, the first color component of the block having a first size, wherein applying the downsampled convolutional neural network layer to the first color component generates a downsampled first color component having a second size smaller than the first size; A first convolutional neural network filter is used to filter the second color component having the second size to form a filtered second color component, wherein the first convolutional neural network filter is different from the downsampled convolutional neural network layer; Wherein, the first color component is a luminance component, and the second color component is either a blue hue chromaticity component or a red hue chromaticity component; The downsampled first color component and the filtered second color component are cascaded to form cascaded color components; and The cascaded color components are filtered using at least a second convolutional neural network filter to form filtered cascaded components, the filtered cascaded components including a filtered downsampled first color component.

2. The method of claim 1, further comprising upsampling the filtered, downsampled first color component to the first size.

3. The method of claim 1, further comprising combining two or more filtered downsampled blocks of the first color component to generate an upsampled first color component having the first size, wherein the first color component includes the filtered downsampled first color component.

4. The method of claim 3, wherein the first size comprises 2N×2N, and wherein the two or more filtered downsampled first color components of the first color component comprise four N×N filtered downsampled blocks of the first color component.

5. The method according to claim 1, wherein the downsampling convolutional neural network layer comprises a 3×3×M convolutional neural network layer filter with a stride of 2.

6. The method of claim 1, wherein the second convolutional neural network filter includes a residual processing unit.

7. The method according to claim 6, wherein the residual processing unit comprises a first 3×3×K×K convolutional layer, a PReLU layer, and a second 3×3×K×K convolutional layer.

8. The method of claim 1, further comprising filtering the third color component of the video data block using a convolutional neural network layer.

9. The method of claim 8, wherein the second dimension comprises a smaller of the dimension of the second color component or the dimension of the third color component.

10. An apparatus for decoding video data, the apparatus comprising: At least one memory, including instructions and configured to store video data; as well as One or more processors, implemented in a circuit and configured to execute the instructions to cause the device to: A downsampled convolutional neural network layer is applied to a first color component of a video data block, the first color component of the block having a first size, wherein applying the downsampled convolutional neural network layer to the first color component generates a downsampled first color component having a second size smaller than the first size; A first convolutional neural network filter is used to filter the second color component having the second size to form a filtered second color component, wherein the first convolutional neural network filter is different from the downsampled convolutional neural network layer; Wherein, the first color component is a luminance component, and the second color component is either a blue hue chromaticity component or a red hue chromaticity component; The downsampled first color component and the filtered second color component are cascaded to form cascaded color components; as well as The cascaded color components are filtered using at least a second convolutional neural network filter to form cascaded color components, the cascaded color components including a filtered and downsampled first color component.

11. The device of claim 10, wherein the one or more processors are further configured to execute the instructions to cause the device to upsample the filtered, downsampled first color component to the first size.

12. The device of claim 10, wherein the one or more processors are further configured to execute the instructions to cause the device to combine two or more filtered downsampled blocks of the first color component to generate an upsampled first color component having the first size, wherein the first color component includes the filtered downsampled first color component.

13. The device of claim 12, wherein the first size comprises 2N×2N, and wherein the two or more filtered downsampled blocks of the first color component comprise four N×N filtered downsampled blocks of the first color component.

14. The device of claim 10, wherein the downsampling convolutional neural network layer comprises a 3×3×M convolutional neural network layer filter with a stride of 2.

15. The apparatus of claim 10, wherein the second convolutional neural network filter includes a residual processing unit.

16. The apparatus of claim 15, wherein the residual processing unit comprises a first 3×3×K×K convolutional layer, a PReLU layer, and a second 3×3×K×K convolutional layer.

17. The device of claim 10, wherein the one or more processors are further configured to execute the instructions to cause the device to filter the third color component of the video data block using the convolutional neural network layer.

18. The device of claim 17, wherein the second dimension includes the smaller of the dimension of the second color component or the dimension of the third color component.

19. The device of claim 10, further comprising a display configured to display video data corresponding to the cascaded color components.

20. The device of claim 10, wherein the device comprises one or more of a camera, a computer, a mobile device, a broadcast receiver device, or a set-top box.

21. A non-transitory computer-readable storage medium having instructions stored thereon, the instructions, when executed, causing a processor to: A downsampled convolutional neural network layer is applied to a first color component of a video data block, the first color component of the block having a first size, wherein applying the downsampled convolutional neural network layer to the first color component generates a downsampled first color component having a second size smaller than the first size; A first convolutional neural network filter is used to filter the second color component having the second size to form a filtered second color component, wherein the first convolutional neural network filter is different from the downsampled convolutional neural network layer; in, The first color component is a luminance component, and the second color component is either a blue hue chromaticity component or a red hue chromaticity component; The downsampled first color component and the filtered second color component are cascaded to form cascaded color components; as well as The cascaded color components are filtered using at least a second convolutional neural network filter to form filtered cascaded components, the filtered cascaded components including a filtered downsampled first color component.

22. The computer-readable storage medium of claim 21, further comprising instructions that cause the processor to upsample the filtered, downsampled first color component to the first size.

23. The computer-readable storage medium of claim 21, further comprising instructions that cause the processor to combine two or more filtered downsampled blocks of the first color component to generate an upsampled first color component having the first size, wherein the first color component includes the filtered downsampled first color component.

24. The computer-readable storage medium of claim 23, wherein the first size comprises 2N×2N, and wherein the two or more filtered downsampled blocks of the first color component comprise four N×N filtered downsampled blocks of the first color component.

25. The computer-readable storage medium of claim 21, wherein the downsampling convolutional neural network layer comprises a 3×3×M convolutional neural network layer filter with a stride of 2.

26. The computer-readable storage medium of claim 21, wherein the second convolutional neural network filter includes a residual processing unit, and the residual processing unit includes a first 3×3×K×K convolutional layer, a PReLU layer, and a second 3×3×K×K convolutional layer.

27. The computer-readable storage medium of claim 21, further comprising instructions for causing the processor to filter a third color component of the video data block using the convolutional neural network layer.

28. The computer-readable storage medium of claim 27, wherein the second size includes a smaller of the size of the second color component or the size of the third color component.

29. An apparatus for filtering and decoding video data, the apparatus comprising: Components for applying a downsampled convolutional neural network layer to a first color component of a video data block, the first color component of the block having a first size, wherein the downsampled convolutional neural network layer is applied to the first color component to generate a downsampled first color component having a second size smaller than the first size; A component for filtering a second color component having the second size using a first convolutional neural network filter to form a filtered second color component, wherein the first convolutional neural network filter is different from the downsampled convolutional neural network layer; Wherein, the first color component is a luminance component, and the second color component is either a blue hue chromaticity component or a red hue chromaticity component; A component for cascading the downsampled first color component and the filtered second color component to form cascaded color components; as well as A component for filtering the cascaded color components using at least a second convolutional neural network filter.

Images