Activation Function Design in Neural Network-Based Filtering Processes for Video Coding

Abstract

1. A method of coding video data, the method comprising: reconstructing a block of the video data; and applying a convolutional neural network (CNN)-based filter to the reconstructed block, wherein the CNN-based filter uses a LeakyReLU activation function.

Description

[Technical Field]

[0001]

[0001] This application claims priority to U.S. Patent Application No. 17 / 489,459, filed September 29, 2021, and U.S. Provisional Patent Application No. 63 / 085,936, filed September 30, 2020, the entire contents of each of which are incorporated by reference. U.S. Patent Application No. 17 / 489,459, filed September 29, 2021, claims the benefit of U.S. Provisional Patent Application No. 63 / 085,936, filed September 30, 2020.

[0002] FIELD OF THE DISCLOSURE

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

[0003] Digital video capabilities may be incorporated into a wide range of devices, including digital televisions, digital direct broadcast systems, wireless broadcast systems, personal digital assistants (PDAs), laptop or desktop computers, tablet computers, e-book readers, digital cameras, digital recording devices, digital media players, video gaming devices, video game consoles, cellular or satellite radiotelephones, so-called "smartphones," video teleconferencing devices, video streaming devices, etc. Digital video devices implement video coding techniques, such as those described in standards defined by MPEG-2, MPEG-4, ITU-T H.263, ITU-T H.264 / MPEG-4, Part 10, Advanced Video Coding (AVC), ITU-T H.265 / High Efficiency Video Coding (HEVC), and extensions to such standards. By implementing such video coding techniques, video devices may more efficiently transmit, receive, encode, decode, and / or store digital video information.

[0004]

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

[0005]

[0005] Generally, this disclosure describes techniques related to a filtering process for distorted pictures. The filtering process may be based on neural network technology. For example, the filtering process may apply a convolutional neural network (CNN)-based filter to a reconstructed block of video data. The CNN-based filter may use a leaky rectified linear unit (ReLU) activation function. The use of the leaky ReLU activation function may lead to better performance during training than a rectified linear unit (ReLU) activation function and to greater stability than a parametric rectified linear unit (PReLU) activation function. The filtering process may be used in the context of advanced video codecs, such as extensions of generic video coding (VVC) or next-generation video coding standards, and any other video codecs.

[0006]

[0006] In one example, the present disclosure describes a method for coding video data, the method comprising reconstructing a block of the video data and applying a convolutional neural network (CNN)-based filter to the reconstructed block, wherein the CNN-based filter uses a LeakyReLU activation function.

[0007]

[0007] In another example, a device for coding video data includes a memory for storing the video data and one or more processors implemented in a circuit, the one or more processors configured to reconstruct blocks of the video data and apply a convolutional neural network (CNN)-based filter to the reconstructed blocks, wherein the CNN-based filter uses a LeakyReLU activation function.

[0008]

[0008] In another example, the present disclosure describes a device for coding video data, the device comprising: means for reconstructing a block of the video data; and means for applying a convolutional neural network (CNN)-based filter to the reconstructed block, wherein the CNN-based filter uses a LeakyReLU activation function.

[0009]

[0009] In another example, the present disclosure describes a computer-readable storage medium having stored thereon instructions that, when executed, cause one or more processors to reconstruct blocks of video data and apply a convolutional neural network (CNN)-based filter to the reconstructed blocks, wherein the CNN-based filter uses a LeakyReLU activation function.

[0010]

[0010] The details of one or more examples are set forth in the accompanying drawings and the description below. Other features, objects, and advantages will become apparent from the description, drawings, and claims. [Brief explanation of the drawings]

[0011] [Figure 1]

[0011] FIG. 1 is a block diagram illustrating an example video encoding and decoding system that may implement the techniques of this disclosure. [Figure 2A]

[0012] 1 is a conceptual diagram illustrating an exemplary quad-tree binary tree (QTBT) structure. [Figure 2B] A conceptual diagram showing the corresponding coding tree unit (CTU). [Figure 3]

[0013] 1 is a block diagram illustrating an example hybrid video coding framework. [Figure 4]

[0014] 1 is a conceptual diagram illustrating an example hierarchical prediction structure with a group of pictures (GOP) size equal to 16. FIG. [Figure 5]

[0015] Conceptual diagram showing a four-layer convolutional neural network (CNN)-based filter. [Figure 6A]

[0016] A conceptual diagram showing the ReLU activation function. [Figure 6B]

[0017] A conceptual diagram showing the PReLU activation function. [Figure 7]

[0018] FIG. 1 is a conceptual diagram illustrating an example CNN-based filter with 14 layers and LeakyReLU as an activation function, in accordance with one or more techniques of this disclosure. [Figure 8]

[0019] 1 is a block diagram illustrating an example video encoder that may implement the techniques of this disclosure. [Figure 9]

[0020] 1 is a block diagram illustrating an example video decoder that may implement the techniques of this disclosure. [Figure 10]

[0021] 10 is a flowchart illustrating an example method for encoding a current block in accordance with techniques of this disclosure. [Figure 11]

[0022] 10 is a flowchart illustrating an example method for decoding a current block, in accordance with techniques of this disclosure. [Figure 12]

[0023] 10 is a flowchart illustrating an example operation of a video coder, in accordance with one or more techniques of this disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0012]

[0024] Video encoders and video decoders may implement in-loop filters that can improve the quality of the reconstructed picture. Common types of in-loop filters include deblocking filters and adaptive loop filters (ALFs). Recently, neural network (NN)-based filters have been proposed. NN-based filters include a neural network that takes blocks of the reconstructed picture as input. The neural network outputs filtered blocks.

[0013]

[0025] A neural network-based filter may include an input layer, one or more hidden layers, and an output layer. These layers may include a set of neurons that receive inputs and generate outputs. Each of the input and hidden layers is associated with an activation function that is applied to the outputs of the neurons in the layer. Rectified Linear Unit (ReLU) and Parametric ReLU (PReLU) are examples of activation functions. The activation function used in a neural network can significantly affect the performance and trainability of the neural network. For example, PReLU may have better performance than ReLU. However, PReLU may destabilize the training process of a neural network because PReLU itself includes trainable variables that can be extreme.

[0014]

[0026] This disclosure describes techniques that may address these technical issues with previously proposed NN-based filters in video coding. For example, a video coder (e.g., a video encoder or video decoder) may reconstruct a block of video data. Further, the video coder may apply a convolutional neural network (CNN)-based filter to the reconstructed block. The CNN-based filter uses a LeakyReLU activation function. The use of the LeakyReLU activation function may enable the CNN-based filter to have better performance than PReLU while avoiding instability during the training process. The use of the LeakyReLU activation function, in contrast to other types of activation functions, may result in unexpected and significant improvements in video coding in terms of both performance and stability based on specific characteristics exhibited during filtering of the reconstructed picture.

[0015]

[0027] 1 is a block diagram illustrating an example video encoding and decoding system 100 that may implement techniques of this disclosure. The techniques of this disclosure are generally directed to coding (encoding and / or decoding) video data. Generally, the video data includes some data for processing the video. Thus, the video data may include raw uncoded video, coded video, decoded (e.g., reconstructed) video, and video metadata, such as signaling data.

[0016]

[0028] 1, system 100 includes a source device 102 that, in this example, provides encoded video data to be decoded and displayed by a destination device 116. In particular, source device 102 provides the video data to destination device 116 via a computer-readable medium 110. Source device 102 and destination device 116 may comprise any of a wide range of devices, including desktop computers, notebook (i.e., laptop) computers, mobile devices, tablet computers, set-top boxes, telephone handsets such as smartphones, televisions, cameras, display devices, digital media players, video gaming consoles, video streaming devices, broadcast receiver devices, etc. In some cases, source device 102 and destination device 116 may be equipped for wireless communication and therefore may be referred to as wireless communication devices.

[0017]

[0029] In the example of FIG. 1 , source device 102 includes a video source 104, memory 106, video encoder 200, and output interface 108. Destination device 116 includes an input interface 122, a video decoder 300, memory 120, and a display device 118. According to this disclosure, video encoder 200 of source device 102 and video decoder 300 of destination device 116 may be configured to apply techniques for using neural network-based filters. Thus, source device 102 represents an example of a video encoding device, and 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.

[0018]

[0030] The system 100 shown in FIG. 1 is merely an example. In general, any digital video encoding and / or decoding device may implement techniques for using neural network-based filters. Source device 102 and destination device 116 are merely examples of coding devices, such that source device 102 generates coded video data for transmission to destination device 116. This disclosure refers to a “coding” device as a device that performs coding (encoding and / or decoding) of data. Accordingly, video encoder 200 and video decoder 300 represent examples of coding devices, particularly video encoders and video decoders, respectively. In some examples, source device 102 and destination device 116 may operate substantially symmetrically, such that each of source device 102 and destination device 116 includes video encoding and video decoding components. Thus, system 100 may support one-way or two-way video transmission between source device 102 and destination device 116, for example, video streaming, video playback, video broadcasting, or video telephony.

[0019]

[0031] Generally, video source 104 represents a source of video data (i.e., raw, unencoded video data) and provides a continuous series of pictures (also called “frames”) of the video data to video encoder 200, which encodes the data for the pictures. Video source 104 of source device 102 may include a video capture device, such as a video camera, a video archive containing previously captured live video, and / or a video feed interface for receiving video from a video content provider. As a further alternative, video source 104 may generate computer-graphics-based data as source video, or a combination of live, archived, and computer-generated video. In each case, video encoder 200 encodes the captured, pre-captured, or computer-generated video data. Video encoder 200 may rearrange the pictures from their received order (sometimes referred to as “display order”) into a coding order for coding. Video encoder 200 may generate a bitstream including the encoded video data. Source device 102 may then output the encoded video data onto computer-readable medium 110 via output interface 108 for receipt and / or retrieval by, for example, input interface 122 of destination device 116.

[0020]

[0032] Memory 106 of source device 102 and memory 120 of destination device 116 represent general-purpose memory. In some examples, memory 106, 120 may store raw video data, e.g., raw video from video source 104 and raw decoded video data from video decoder 300. Additionally or alternatively, memory 106, 120 may store software instructions executable by video encoder 200 and video decoder 300, respectively, for example. While memory 106 and memory 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. Additionally, memory 106, 120 may store encoded video data, e.g., output from video encoder 200 and input to video decoder 300. In some examples, portions of the memory 106, 120 may be allocated as one or more video buffers, for example, to store raw decoded and / or encoded video data.

[0021]

[0033] The computer-readable medium 110 may represent any type of medium or device capable of transporting encoded video data from the source device 102 to the destination device 116. In one example, the computer-readable medium 110 represents a communication medium for enabling the source device 102 to transmit the encoded video data directly to the destination device 116 in real time, for example, via a radio frequency network or a computer-based network. The output interface 108 may modulate a transmission signal including the encoded video data, and the input interface 122 may demodulate a received transmission signal in accordance with a communication standard such as a wireless communication protocol. The communication medium may comprise any wireless or wired communication medium, such as a radio frequency (RF) spectrum or one or more physical transmission lines. The communication medium may form part of a packet-based network, such as a local area network, a wide area network, or a global network such as the Internet. The communication medium may include routers, switches, base stations, or any other equipment that may be useful in facilitating communication from the source device 102 to the destination device 116.

[0022]

[0034] In some examples, source device 102 may output the encoded data from output interface 108 to storage device 112. Similarly, destination device 116 may access the encoded data from storage device 112 via input interface 122. Storage device 112 may include any of a variety of distributed or locally accessed data storage media, such as a hard drive, Blu-ray disc, DVD, CD-ROM, flash memory, volatile or non-volatile memory, or any other suitable digital storage medium for storing encoded video data.

[0023]

[0035] In some examples, source device 102 may output the encoded video data to file server 114 or another intermediate storage device, which may store the encoded video data generated by source device 102. Destination device 116 may access the stored video data from file server 114 via streaming or download.

[0024]

[0036] The file server 114 may be any type of server device capable of storing encoded video data and transmitting the encoded video data to the destination device 116. The file server 114 may represent a web server (e.g., for a website), a server configured to provide file transfer protocol services (e.g., File Transfer Protocol (FTP) or File Delivery over Unidirectional Transport (FLUTE) 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. The file server 114 may additionally or alternatively implement one or more HTTP streaming protocols, such as Dynamic Adaptive Streaming over HTTP (DASH), HTTP Live Streaming (HLS), Real Time Streaming Protocol (RTSP), HTTP Dynamic Streaming, etc.

[0025]

[0037] Destination device 116 may access the encoded video data from file server 114 through any standard data connection, including an Internet connection. This may include a wireless channel (e.g., a Wi-Fi connection), a wired connection (e.g., a digital subscriber line (DSL), a cable modem, etc.), or a combination of both, that is suitable for accessing the encoded video data stored on file server 114. Input interface 122 may be configured to operate according to any one or more of the various protocols described above for retrieving or receiving media data from file server 114, or other such protocols for retrieving media data.

[0026]

[0038] Output interface 108 and input interface 122 may represent a wireless transmitter / receiver, a modem, a wired networking component (e.g., an Ethernet card), a wireless communication component operating according to any of the various IEEE 802.11 standards, or other physical components. In examples in which output interface 108 and input interface 122 comprise wireless components, output interface 108 and input interface 122 may be configured to transfer data, such as encoded video data, according to a cellular communication standard, such as 4G, 4G-LTE (Long Term Evolution), LTE-Advanced, 5G, etc. In some examples in which output interface 108 comprises a wireless transmitter, output interface 108 and input interface 122 may be configured to transfer data, such as encoded video data, according to other wireless standards, such as the IEEE 802.11 specification, the IEEE 802.15 specification (e.g., ZigBee), the Bluetooth standard, etc. In some examples, source device 102 and / or destination device 116 may include respective system-on-chip (SoC) devices. For example, the source device 102 may include an SoC device for performing functions attributed to the video encoder 200 and / or the output interface 108, and the destination device 116 may include an SoC device for performing functions attributed to the video decoder 300 and / or the input interface 122.

[0027]

[0039] The techniques of this disclosure may be applied to video coding supporting any of a variety of multimedia applications, such as over-the-air television broadcast, cable television transmission, satellite television transmission, Internet streaming video transmission such as Dynamic Adaptive Streaming over HTTP (DASH), digital video encoded on a data storage medium, decoding of digital video stored on a data storage medium, or other applications.

[0028]

[0040] The input interface 122 of the destination device 116 receives the encoded video bitstream from the computer-readable medium 110 (e.g., a communications medium, a storage device 112, a file server 114, etc.). The encoded video bitstream may include signaling information defined by the video encoder 200 that is also used by the video decoder 300, such as syntax elements having values ​​that describe characteristics and / or processing of video blocks or other coded units (e.g., slices, pictures, groups of pictures, sequences, etc.). The display device 118 displays decoded pictures of the decoded video data to a user. The display device 118 may represent any of a variety of display devices, such as a liquid crystal display (LCD), a plasma display, an organic light-emitting diode (OLED) display, or another type of display device.

[0029]

[0041] 1, in some examples, video encoder 200 and video decoder 300 may each be integrated with an audio encoder and / or decoder and may include an appropriate MUX-DEMUX unit 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 the User Datagram Protocol (UDP).

[0030]

[0042] The video encoder 200 and the video decoder 300 may each be implemented as any of a variety of suitable encoder and / or decoder circuits, or any combination thereof, 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, etc. When the techniques are implemented partially in software, a device may store instructions for the software on a suitable non-transitory computer-readable medium and execute the instructions in hardware using one or more processors to implement the techniques of this disclosure. Each of the video encoder 200 and the video decoder 300 may be included in one or more encoders or decoders, any of which may be integrated as part of a combined encoder / decoder (codec) in the respective device. Devices including the video encoder 200 and / or the video decoder 300 may comprise integrated circuits, microprocessors, and / or wireless communication devices such as cellular phones.

[0031]

[0043] Video encoder 200 and video decoder 300 may operate according to a video coding standard, such as ITU-T H.265, also known as High Efficiency Video Coding (HEVC), or extensions thereof, such as multiview and / or scalable video coding extensions. Alternatively, video encoder 200 and video decoder 300 may operate according to other proprietary or industry standards, such as ITU-T H.266, also known as Generic Video Coding (VVC). Version 1 of the VVC specification has been finalized (hereinafter referred to as VVC FDIS) and is available from http: / / phenix.int-evry.fr / jvet / doc_end_user / documents / 19_Teleconference / wg11 / JVET-S2001-v17.zip. However, the techniques of this disclosure are not limited to any particular coding standard.

[0032]

[0044] Generally, the video encoder 200 and the video decoder 300 may perform block-based coding of pictures. The term “block” generally refers to a structure containing data to be processed (e.g., encoded, decoded, or otherwise used in the encoding and / or decoding process). For example, a block may include a two-dimensional matrix of luminance and / or chrominance data samples. Generally, the video encoder 200 and the video decoder 300 may code video data represented in YUV (e.g., Y, Cb, Cr) format. That is, rather than coding red, green, and blue (RGB) data for a picture sample, the video encoder 200 and the video decoder 300 may code a luminance component and a chrominance component, where the chrominance component may include both red and blue hues of chrominance components. In some examples, the video encoder 200 converts received RGB-formatted data to a YUV representation prior to encoding, and the video decoder 300 converts the YUV representation to an RGB format. Alternatively, pre-processing and post-processing units (not shown) may perform these conversions.

[0033]

[0045] This disclosure may generally refer to coding (e.g., encoding and decoding) a picture to include the process of encoding or decoding data for a picture. Similarly, this disclosure may refer to coding a block of a picture to include the process of encoding or decoding data for the block, e.g., predictive and / or residual coding. A coded video bitstream generally includes a series of values ​​for syntax elements that represent coding decisions (e.g., coding modes) and the partitioning of a picture into blocks. Thus, references to coding a picture or a block should be understood generally as coding values ​​for the syntax elements that form the picture or block.

[0034]

[0046] HEVC defines various blocks, including coding units (CUs), prediction units (PUs), and transform units (TUs). According to HEVC, a video coder (such as video encoder 200) partitions coding tree units (CTUs) into CUs according to a quadtree structure. That is, the video coder partitions CTUs and CUs into four equal, non-overlapping squares, and each node of the quadtree has either zero or four child nodes. A node without child nodes may be referred to as a "leaf node," and a CU of such a leaf node may include one or more PUs and / or one or more TUs. The video coder may further partition PUs and TUs. For example, in HEVC, a residual quadtree (RQT) represents the partitioning of TUs. In HEVC, a PU represents inter-predicted data, and a TU represents residual data. An intra-predicted CU includes intra-prediction information, such as an intra-mode indication.

[0035]

[0047] As another example, video encoder 200 and video decoder 300 may be configured to operate according to VVC. According to VVC, a video coder (such as video encoder 200) partitions a picture into multiple coding tree units (CTUs). Video encoder 200 may partition the CTUs according to a tree structure, such as a quad-tree binary tree (QTBT) structure or a multi-type tree (MTT) structure. The QTBT structure eliminates the concept of multiple partition types, such as the separation between CUs, PUs, and TUs in HEVC. The QTBT structure includes two levels: a first level partitioned according to quad-tree partitioning and a second level partitioned according to binary tree partitioning. The root node of the QTBT structure corresponds to a CTU. The leaf nodes of the binary tree correspond to coding units (CUs).

[0036]

[0048] In the MTT partitioning structure, blocks may be partitioned using quad tree (QT) partitioning, binary tree (BT) partitioning, and one or more types of triple tree (TT) (also called ternary tree (TT)) partitioning. Triple tree or ternary tree partitioning is a partitioning in which a block is split into three sub-blocks. In some examples, triple tree or ternary tree partitioning divides a block into three sub-blocks without splitting the original block through the center. The partitioning types in MTT (e.g., QT, BT, and TT) can be symmetric or asymmetric.

[0037]

[0049] 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 luminance and chrominance components, and 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 luminance component and another QTBT / MTT structure for both chrominance components (or two QTBT / MTT structures for each chrominance component).

[0038]

[0050] Video encoder 200 and video decoder 300 may be configured to use quadtree partitioning, QTBT partitioning, MTT partitioning, or other partition structures according to HEVC. For illustrative purposes, the description of the techniques of this disclosure is presented with respect to QTBT partitioning. However, it should be understood that the techniques of this disclosure may also be applied to video coders configured to use quadtree partitioning, or other types of partitioning as well.

[0039]

[0051] In some examples, a CTU includes a coding tree block (CTB) of luma samples, two corresponding CTBs of chroma samples for a picture with three sample arrays, or a CTB of samples for a monochrome picture, or a picture coded using three separate color planes and syntax structures used to code the samples. The CTB may be an N×N block of samples, for some value of N such that the division of the components into the CTB is partitioned. A component is an array or a single sample from one of three arrays (luma and two chroma) that configure the picture in 4:2:0, 4:2:2, or 4:4:4 color format, or an array or a single sample of an array that configures the picture in monochrome format. In some examples, a coding block is an M×N block of samples, for some value of M and N such that the division of the CTB into coding blocks is partitioned.

[0040]

[0052] Blocks (e.g., CTUs or CUs) may be grouped in various ways in a picture. As an example, a brick may refer to a rectangular region of a CTU row within a particular tile in a picture. A tile may be a rectangular region of a CTU within a particular tile column and a particular tile row in a picture. A tile column refers to a rectangular region of a CTU having a height equal to the height of the picture and a width specified by a syntax element (e.g., in a picture parameter set). A tile row refers to a rectangular region of a CTU having a height specified by a syntax element (e.g., in a picture parameter set) and a width equal to the width of the picture.

[0041]

[0053] In some examples, a tile may be partitioned into multiple bricks, each of which may contain one or more CTU rows within the tile. A tile that is not partitioned into multiple bricks may also be referred to as a brick. However, a brick that is a true subset of a tile may not be referred to as a tile.

[0042]

[0054] The bricks in a picture may also be arranged into slices. A slice may be an integer number of bricks of a picture that may be contained entirely in a single Network Abstraction Layer (NAL) unit. In some examples, a slice includes either several complete tiles or only a continuous sequence of complete bricks of one tile.

[0043]

[0055] This disclosure may use "N x N" and "N by N" interchangeably to refer to the sample dimensions of a block (such as a CU or other video block) in terms of the vertical and horizontal dimensions, e.g., 16 x 16 samples or 16 by 16 samples. Generally, a 16 x 16 CU has 16 samples in the vertical direction (y = 16) and 16 samples in the horizontal direction (x = 16). Similarly, an N x N CU generally 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 may be arranged in rows and columns. Moreover, a CU does not necessarily have to have the same number of samples in the horizontal direction as in the vertical direction. For example, a CU may comprise N x M samples, where M is not necessarily equal to N.

[0044]

[0056] Video encoder 200 encodes video data for a CU, which represents prediction and / or residual information and other information. The prediction information indicates how the CU should be predicted to form a predictive block for the CU. The residual information generally represents sample-by-sample differences between the samples of the CU prior to encoding and the predictive block.

[0045]

[0057] To predict a CU, the video encoder 200 may generally form a predictive block for the CU through inter-prediction or intra-prediction. Inter-prediction generally refers to predicting a CU from data of a previously coded picture, while intra-prediction generally refers to predicting a CU from previously coded data of the same picture. To perform inter-prediction, the video encoder 200 may generate a predictive block using one or more motion vectors. The video encoder 200 may generally perform motion search to identify a reference block that closely matches the CU, e.g., with respect to the difference between the CU and the reference block. The video encoder 200 may calculate a difference metric using a sum of absolute differences (SAD), a sum of squared differences (SSD), a mean absolute difference (MAD), a mean squared difference (MSD), or other such difference calculation to determine whether the reference block closely matches the current CU. In some examples, the video encoder 200 may predict the current CU using unidirectional prediction or bidirectional prediction.

[0046]

[0058] Some examples of VVC also provide an affine motion compensation mode, which may be considered an inter-prediction mode. In an affine motion compensation mode, video encoder 200 may determine two or more motion vectors that represent non-translational motion, such as zooming in or out, rotation, perspective motion, or other irregular motion types.

[0047]

[0059] To perform intra prediction, video encoder 200 may select an intra prediction mode to generate a predictive block. Some examples of VVC provide 67 intra prediction modes, including various directional modes as well as planar and DC modes. Generally, video encoder 200 selects an intra prediction mode that describes neighboring samples relative to a current block (e.g., a block of a CU) from which samples of the current block should be predicted. Such samples may generally be above, above-left, or to the left of the current block in the same picture as the current block, assuming that video encoder 200 codes CTUs and CUs in raster scan order (left to right, top to bottom).

[0048]

[0060] Video encoder 200 encodes data representing a prediction mode for the current block. For example, in an inter prediction mode, video encoder 200 may encode data representing which of various available inter prediction modes is used, as well as motion information for the corresponding mode. For example, in unidirectional or bidirectional inter prediction, video encoder 200 may encode motion vectors using advanced motion vector prediction (AMVP) or merge mode. Video encoder 200 may use similar modes to encode motion vectors for affine motion compensation modes.

[0049]

[0061] Following prediction, such as intra-prediction or inter-prediction, of a block, the video encoder 200 may calculate residual data for the block. The residual data, such as a residual block, represents sample-by-sample differences between the block and a predictive block for the block formed using a corresponding prediction mode. The video encoder 200 may apply one or more transforms to the residual block to produce transformed data in a transform domain rather than the sample domain. For example, the video encoder 200 may apply a discrete cosine transform (DCT), an integer transform, a wavelet transform, or a conceptually similar transform to the residual video data. Furthermore, the video encoder 200 may apply a secondary transform, such as a mode-dependent non-separable secondary transform (MDNSST), a signal-dependent transform, or a Karhunen-Loeve transform (KLT), following the first transform. The video encoder 200 produces transform coefficients following application of the one or more transforms.

[0050]

[0062] As described above, following any transformation to produce transform coefficients, the video encoder 200 may perform quantization of the transform coefficients. Quantization generally refers to a process in which transform coefficients are quantized to possibly reduce the amount of data used to represent the transform coefficients, thereby providing further compression. By performing the quantization process, the video encoder 200 may reduce the bit depth associated with some or all of the transform coefficients. For example, the video encoder 200 may round an n-bit value to an m-bit value during quantization, where n is greater than m. In some examples, to perform quantization, the video encoder 200 may perform a bitwise right shift of the value to be quantized.

[0051]

[0063] Following quantization, the video encoder 200 may scan the transform coefficients to create a one-dimensional vector from a two-dimensional matrix including the quantized transform coefficients. The scan may be designed to place higher energy (and therefore lower frequency) transform coefficients at the front of the vector and lower energy (and therefore higher frequency) transform coefficients at the back of the vector. In some examples, the video encoder 200 may utilize a predefined scan order to scan the quantized transform coefficients to create a serialized vector and then entropy code the quantized transform coefficients of the vector. In other examples, the video encoder 200 may perform adaptive scanning. After scanning the quantized transform coefficients to form the one-dimensional vector, the video encoder 200 may entropy code the one-dimensional vector, for example, according to context-adaptive binary arithmetic coding (CABAC). The video encoder 200 may also entropy code values ​​for syntax elements describing metadata associated with the encoded video data for use by the video decoder 300 in decoding the video data.

[0052]

[0064] To implement CABAC, video encoder 200 may assign a context in a context model to a symbol to be transmitted. The context may relate, for example, to whether neighboring values ​​of the symbol are zero values. A probability determination may be based on the context assigned to the symbol.

[0053]

[0065] Video encoder 200 may further generate syntax data, such as block-based syntax data, picture-based syntax data, and sequence-based syntax data, for video decoder 300, e.g., in a picture header, a block header, a slice header, or other syntax data, such as a sequence parameter set (SPS), a picture parameter set (PPS), or a video parameter set (VPS). Video decoder 300 may similarly decode such syntax data to determine how to decode corresponding video data.

[0054]

[0066] In this manner, video encoder 200 may generate a bitstream including syntax elements that describe encoded video data, e.g., partitions of a picture into blocks (e.g., CUs) and prediction and / or residual information for the blocks. Finally, video decoder 300 may receive the bitstream and decode the encoded video data.

[0055]

[0067] Generally, video decoder 300 performs an inverse process to that performed by video encoder 200 to decode encoded video data of a bitstream. For example, video decoder 300 may decode values ​​for syntax elements of a bitstream using CABAC in a manner that is reverse to, but substantially similar to, the CABAC encoding process of video encoder 200. The syntax elements may define partition information for partitioning a picture into CTUs and the partitioning of each CTU according to a corresponding partition structure, such as a QTBT structure, to define CUs of the CTU. The syntax elements may further define prediction and residual information for blocks of video data (e.g., CUs).

[0056]

[0068] The residual information may be represented, for example, by quantized transform coefficients. The video decoder 300 may dequantize and inverse transform the quantized transform coefficients of the block to reconstruct a residual block for the block. The video decoder 300 uses the signaled prediction mode (intra- or inter-prediction) and associated prediction information (e.g., motion information for inter-prediction) to form a predictive block for the block. The video decoder 300 may then combine the predictive block and the residual block (sample by sample) to reconstruct the original block. The video decoder 300 may perform additional processing, such as performing a deblocking process to reduce visual artifacts along block boundaries.

[0057]

[0069] This disclosure may generally refer to “signaling” certain information, such as syntax elements. The term “signaling” may generally refer to communication of values ​​for syntax elements and / or other data used to decode encoded video data. That is, video encoder 200 may signal values ​​for syntax elements in a bitstream. Generally, signaling refers to generating values ​​in a bitstream. As mentioned above, source device 102 may transport the bitstream to destination device 116 in substantially real time or may transport the bitstream to destination device 116 in non-real time, such as may be done when storing syntax elements to storage device 112 for later retrieval by destination device 116.

[0058]

[0070] 2A and 2B are conceptual diagrams illustrating an exemplary quad-tree binary tree (QTBT) structure 130 and a corresponding coding tree unit (CTU) 132. Solid lines represent quad-tree splitting, and dotted lines indicate binary-tree splitting. At each split (i.e., non-leaf) node of the binary tree, one flag is signaled to indicate which splitting type (i.e., horizontal or vertical) is used, where, in this example, 0 indicates horizontal splitting and 1 indicates vertical splitting. In quad-tree splitting, the quad-tree node splits a block horizontally and vertically into four sub-blocks of equal size, so there is no need to indicate the splitting type. Thus, video encoder 200 may encode, and video decoder 300 may decode, syntax elements (such as splitting information) for the region tree level (i.e., solid lines) of QTBT structure 130 and syntax elements (such as splitting information) for the prediction tree level (i.e., dashed lines) of QTBT structure 130. Video encoder 200 may encode, and video decoder 300 may decode, video data, such as prediction and transform data, for CUs represented by terminal leaf nodes of QTBT structure 130.

[0059]

[0071] 2B may be associated with parameters that define the sizes of blocks corresponding to nodes of the QTBT structure 130 at the first and second levels. These parameters may include a CTU size (representing the size of the CTU 132 in the sample), a minimum quadtree size (MinQTSize, representing the minimum allowable quadtree leaf node size), a maximum binary tree size (MaxBTSize, representing the maximum allowable binary tree root node size), a maximum binary tree depth (MaxBTDepth, representing the maximum allowable binary tree depth), and a minimum binary tree size (MinBTSize, representing the minimum allowable binary tree leaf node size).

[0060]

[0072] The root node of the QTBT structure corresponding to a CTU may have four child nodes at the first level of the QTBT structure, each of which may be partitioned according to quadtree partitioning. That is, a node at the first level is either a leaf node (with no child nodes) or has four child nodes. The example QTBT structure 130 represents a node including a parent node and child nodes with solid lines for branching. If a node at the first level is not larger than the maximum allowable binary tree root node size (MaxBTSize), the node may be further partitioned by its respective binary tree. The binary tree splitting of a node may be repeated until the node resulting from the split reaches the minimum allowable binary tree leaf node size (MinBTSize) or the maximum allowable binary tree depth (MaxBTDepth). The example QTBT structure 130 represents a node with dashed lines for branching. The binary tree leaf nodes are called coding units (CUs), and CUs are used for prediction (e.g., intra-picture or inter-picture prediction) and transform without further partitioning. As explained above, a CU may also be referred to as a "video block" or "block."

[0061]

[0073] In one example of a QTBT partitioning structure, the CTU size is set as 128x128 (luma sample and two corresponding 64x64 chroma samples), MinQTSize is set as 16x16, MaxBTSize is set as 64x64, MinBTSize (for both width and height) is set as 4, and MaxBTDepth is set as 4. Quad-tree partitioning is first applied to the CTU to generate a quad-tree leaf node. The quad-tree leaf node may have a size from 16x16 (i.e., MinQTSize) to 128x128 (i.e., CTU size). If the quad-tree leaf node is 128x128, the leaf quad-tree node is not further split by the binary tree because its size exceeds MaxBTSize (i.e., 64x64 in this example). Otherwise, the quad-tree leaf node is further partitioned by the binary tree. Therefore, the quadtree leaf node is also the root node for the binary tree and has a binary tree depth of 0. When the binary tree depth reaches MaxBTDepth (4 in this example), no further splitting is allowed. A binary tree node with a width equal to MinBTSize (4 in this example) implies that no further vertical splitting (i.e., width division) is allowed for that binary tree node. Similarly, a binary tree node with a height equal to MinBTSize implies that no further horizontal splitting (i.e., height division) is allowed for that binary tree node. As mentioned above, the leaf nodes of the binary tree are called CUs and are further processed according to the prediction and transformation without further division.

[0062]

[0074] Most video coding standards, starting with H.261, are based on the so-called hybrid video coding principle, shown in Figure 3. The term hybrid refers to the combination of two means for reducing redundancy in a video signal: prediction and transform coding with quantization of the prediction residual. While prediction and transform reduce redundancy in a video signal by decorrelation, quantization reduces the data of the transform coefficient representation by reducing their precision, ideally removing only irrelevant details. This hybrid video coding design principle is also used in two recent standards, HEVC and VVC.

[0063]

[0075] A modern hybrid video coder may be composed of the following building blocks, as shown in the example of FIG. 3. In other words, FIG. 3 is a block diagram illustrating an exemplary hybrid video coding framework 320. In the example of FIG. 3, the framework 320 receives a picture 322 of input video data and splits the picture 322 into blocks. Block partitioning is used to divide an image into smaller blocks for the operation of prediction and transform processes. Early video coding standards used a fixed block size, typically 16×16 samples. More recent standards, such as HEVC and VVC, employ a tree-based partitioning structure to provide flexible partitioning, as shown in, for example, FIGS. 2A and 2B.

[0064]

[0076] The framework 320 includes a motion estimation unit 324, an intra prediction unit 326, and an inter prediction unit 328. The motion estimation unit 324 may perform motion estimation based on a previously reconstructed picture 330 of the video data. The inter prediction unit 328 may use motion information generated by the motion estimation unit 324 to generate a prediction block. Motion-compensated prediction or inter-picture prediction (i.e., inter prediction) exploits redundancy that exists between pictures (hence "inter") of a video sequence. In block-based motion compensation used in modern video codecs, prediction is achieved from one or more previously decoded pictures, i.e., reference picture(s). The corresponding area for generating the inter prediction is indicated by motion information, including a motion vector and a reference picture index.

[0065]

[0077] The intra prediction unit 326 may use intra prediction to generate a predictive block. Intra-picture prediction exploits spatial redundancy that exists within a picture (hence "intra") by deriving a prediction for a block from spatially neighboring (reference) samples that have already been coded / decoded. Directional angle prediction, DC prediction, and plane or planar prediction are used in modern video codecs, including AVC, HEVC, and VVC. The framework 320 may select between the predictive block generated by the inter prediction unit 328 and the predictive block generated by the intra prediction unit 326.

[0066]

[0078] A residual generation unit 332 of the framework 320 may generate residual data based on the prediction block and the corresponding block of the picture 322. A transform unit 334 may apply a transform to the residual data to generate transform coefficients. Hybrid video coding standards apply block transforms to the prediction residual (regardless of whether the prediction residual comes from inter-picture prediction or intra-picture prediction). Early standards, including H.261 / 262 / 263, employ the discrete cosine transform (DCT). In HEVC and VVC, more transform kernels besides the DCT may be applied to account for different statistics in a particular video signal.

[0067]

[0079] A quantization unit 336 may then apply quantization to the transform coefficients. Quantization aims to reduce the precision of an input value or set of input values ​​in order to reduce the amount of data needed to represent those values. In hybrid video coding, quantization is generally applied to individual transformed residual samples, i.e., transform coefficients, resulting in integer coefficient levels. In recent video coding standards, the step size used in quantization is derived from a so-called quantization parameter (QP), which controls fidelity and bit rate. A larger step size reduces the bit rate but also degrades quality, which may cause video pictures to exhibit blocking artifacts and blurred details.

[0068]

[0080] An entropy coding unit 338 may apply entropy coding to syntax elements representing transform coefficients. The framework 320 may output the entropy-coded syntax elements in a bitstream. Entropy coding techniques such as context-adaptive binary arithmetic coding (CABAC) may be used in recent video codecs, such as AVC, HEVC, and VVC, due to their high efficiency.

[0069]

[0081] Further, in the example of FIG. 3, an inverse quantization unit 340 may inverse quantize the transform coefficients. An inverse transform unit 342 may apply an inverse transform to the dequantized transform coefficients to generate residual data. A reconstruction unit 344 may reconstruct a block based on the predictive block and the residual data. A loop filter unit 346 may apply one or more filters to a picture including the reconstructed block.

[0070]

[0082] In recent video codecs, a hierarchical prediction structure within a group of pictures (GOP) is applied to improve coding efficiency. FIG. 4 is a conceptual diagram illustrating an exemplary hierarchical prediction structure 400 for a group of pictures (GOP) size equal to 16. In the example of FIG. 4, picture I0 is an intra picture that is not coded using inter prediction. The other pictures in FIG. 4 (i.e., pictures B1-B16) may be coded using inter prediction. In particular, each arrow in FIG. 4 pointing from a first picture to a second picture indicates that the second picture may be coded using the first picture.

[0071]

[0083] Post-loop / in-loop filtering is a filtering process (or a combination of such processes) applied to a reconstructed picture to reduce coding artifacts. The input of the filtering process is generally a reconstructed picture, which is a combination of a reconstructed residual signal (including quantization error) and a prediction. As shown in FIG. 3, the reconstructed picture 330 after in-loop filtering is stored and used as a reference for inter-picture prediction of a later coded picture. Coding artifacts are largely determined by QP, and therefore QP information is generally used in the design of the filtering process. In HEVC, the in-loop filter includes deblocking filtering and sample adaptive offset (SAO) filtering. In VVC, an adaptive loop filter (ALF) was introduced as the third filter. The ALF filtering process is shown below.

[0072]

number

[0073] In equation (1), R(i,j) is the sample before the application of the filtering process, R'(i,j) is the sample value after the application of the filtering process, f(k,l) denotes the filter coefficient, K(x,y) is the clipping function, and c(k,l) denotes the clipping parameter. The variables k and l are

[0074]

number

[0075] from

[0076]

number

[0077] where L denotes the filter length. The clipping function K(x,y)=min(y,max(-y,x)), which corresponds to the function Clip3(-y,y,x). The clipping operation introduces nonlinearity to make the ALF more efficient by reducing the influence of neighboring sample values ​​that are too different from the current sample value. In VVC, the filtering parameters can be signaled in the bitstream, or the filtering parameters can be selected from a predefined filter set. The ALF filtering process can also be summarized as the following equation:

[0078]

number

[0079]

[0084] Some studies have shown that embedding neural networks (NNs) into hybrid video coding frameworks can improve compression efficiency. Neural networks have been used in modules for intra-prediction and inter-prediction to improve prediction efficiency. NN-based in-loop filters have also been an active research topic in recent years. In some examples, the filtering process is applied as a post-filter. When the filtering process is applied as a post-filter, the filtering process is applied only to the output picture, and the unfiltered picture is used as the reference picture. In contrast, when the filtering process is applied as an in-loop filter, the filtering process is applied to the reconstructed picture used for output and used as the reference picture.

[0080]

[0085] The NN-based filter may be further applied on top of an existing filter, such as a deblocking filter, an SAO, or an ALF. The NN-based filter may also be applied exclusively, where the NN-based filter is designed to replace all existing filters.

[0081]

[0086] FIG. 5 is a conceptual diagram illustrating a convolutional neural network (CNN)-based filter 500 with four layers. The CNN-based filter 500 is a specific type of NN-based filter that uses convolutional layers. As shown in FIG. 5, the NN-based filtering process may take reconstructed samples as inputs, and the intermediate output 501 is a residual sample that is added to the input to refine the input samples. The CNN-based filter 500 may use all color components (e.g., luma, Cb chroma, Cr chroma) as inputs to exploit cross-component correlation. Different color components may share the same filter (including network structure and model parameters), or each color component may have its own specific filter.

[0082]

[0087] In particular, the input to the CNN-based filter 500 includes multiple sample blocks 502 of a reconstructed picture. Each of the sample blocks 502 has a size N×N. The sample blocks 502 are rectangular arrays of samples. The sample blocks 502 may or may not correspond to coding blocks, prediction blocks, transform blocks, coding tree blocks, or other types of blocks used in the coding process. However, this disclosure is not limited to square sample blocks and may generally apply to rectangular sample blocks having a size N×M. A chroma block has half the number of samples of a luma block in the vertical direction and half the number of samples of a luma block in the horizontal direction. Thus, for each N×N chroma block, the input to the filter 500 may include four N×N luma blocks. In the example of FIG. 5, the luma blocks are shaded and the chroma blocks are white. Thus, a sample block received as input to filter 500 can be represented as Y×4+U+V, where Y corresponds to luma, U corresponds to the first chroma component, and V corresponds to the second chroma component.

[0083]

[0088] Further, the CNN-based filter 500 includes an input layer 504, hidden layers 506A-506B (collectively "hidden layers 506"), and an output layer 508. Specifically, in the example of FIG. 5, the CNN-based filter 500 includes two hidden layers for a total of four layers. The input layer 504, hidden layer 506, and output layer 508 each include one of convolutional layers 510A-510D (collectively "convolutional layers 510"). Further, the input layer 504 and hidden layer 506 each include one of PReLU activation layers 512A-512C (collectively "PReLU activation layers 512"). Each of the PReLU activation layers 512 takes the output of a convolutional layer as input and applies a PReLU activation function to the input.

[0084]

[0089] Throughout this disclosure, the notation a×b×c×d is used to describe a convolutional layer. The values ​​a and b may correspond to the width and height of a two-dimensional array of input locations. Each of the input locations has c input channels. Thus, a convolutional layer receives a×b×c input values. The value d indicates the number of output channels of the convolutional layer. Each output channel corresponds to a different set of trainable weights applied by the convolutional layer. The number of weights in the set of weights corresponding to an output channel is equal to a×b×c. The value of an output channel (i.e., output value) may be calculated by convolving the input value with the set of weights corresponding to the output channel. For example, the value of an output channel may be calculated as the dot product of the input value and the set of weights corresponding to the output channel. Because there are d output channels, a convolutional layer is associated with a×b×c×d weights.

[0085]

[0090] To illustrate this for the example of FIG. 5, convolutional layer 510A is described as 3×3×6×8. Thus, the input of convolutional layer 510A may be a set of six values ​​corresponding to each chroma sample location in a 3×3 square of chroma sample locations centered at the current chroma sample location. Each chroma sample location corresponds to a chroma sample location in block 502. In this example, the chroma block is half the width and half the height of the luma block. Thus, for each chroma sample location, there are four luma sample locations. For each chroma sample location, there are two types of chroma samples, resulting in six sample values ​​for each chroma sample location. Convolutional layer 510A has eight output channels and therefore generates eight output values. In the example of FIG. 5, PReLU activation layer 512A applies a PReLU activation function to the eight output values ​​generated by convolutional layer 510A. Convolutional layer 510A may repeat this operation with all or some of the chroma sample locations of block 502 as the current chroma sample locations. The set of eight values ​​output by PReLU activation layer 512A may be treated as a value corresponding to a single location in a two-dimensional array of locations used as input for convolutional layer 510B. Convolutional layers 510B, 510C, and 510D operate in a similar manner. Convolutional layer 510D has six output channels corresponding to the six sample values ​​(i.e., four luma and two chroma) corresponding to the current chroma sample location. In other examples, convolutional layer 510A may have values ​​of a, b, c, and d different from 3, 3, 6, and 8. In other words, however, the present disclosure is not limited to filter kernels with 3 × 3 chroma sample locations, but may be applied to general a × b filter kernels, including non-rectangular filter kernels. Additionally, padding may be used at block boundaries, CTU boundaries, and / or picture boundaries.

[0086]

[0091] The output layer 508 of the CNN-based filter 500 generates an intermediate output 501. The intermediate output 501 may include residual samples. These residual samples may differ from the residual samples generated by the residual generation unit 332. The CNN-based filter 500 generates a filtered block 514 by adding the residual samples of the intermediate output 501 to corresponding samples of the sample block 502. The filtered block 514 may include two chroma blocks of size N×N and four luma blocks of size N×N.

[0087]

[0092] The filtering process of FIG. 5 can be generalized as follows.

[0088]

number

[0089]

[0093] 6A is a conceptual diagram illustrating a ReLU activation function 600. As shown in FIG. 6A, the ReLU activation function 600 is

[0090]

number

[0091] It can be defined as:

[0092]

[0094] 6B is a conceptual diagram illustrating a PReLU activation function 650. As shown in FIG. 6B, the PReLU activation function 650 is

[0093]

number

[0094] It can be defined as: In Equation 5, a is a value that is trained as part of training a neural network that includes a PReLU activation function.

[0095] The model structure and model parameters of the NN-based filter may be predefined and stored in the video encoder 200 and the video decoder 300. The model structure and model parameters of the NN-based filter may also be signaled in the bitstream (e.g., by the video encoder 200). For example, the video encoder 200 may encode, and the video decoder 300 may decode, one or more syntax elements having values ​​indicating the model structure and model parameters (e.g., of the NN-based filter). The video encoder 200 may encode the syntax elements in any suitable syntax structure, such as a sequence parameter set (SPS), a picture parameter set (PPS), a video parameter set (VPS), a slide header, etc.

[0096]

[0096] The performance of PReLU has been found to be better than ReLU in most cases. However, PReLU can also make the training process unstable. Therefore, training of a NN-based filter using a PReLU activation function may take longer or may diverge from the optimal solution.

[0097]

[0097] In this disclosure, we propose to use LeakyReLU as the activation function in the NN-based filtering process. LeakyReLU is shown in Equation (6) below.

[0098]

number

[0099] In equation (6), the parameter Alpha is a fixed value. In other words, the parameter Alpha does not change during the training process of a neural network that includes a LeakyReLU activation function. In contrast, the parameter a in PReLU is a trainable variable that can change during the training process.

[0100] As an example, a CNN-based filter is proposed, as shown in the example of FIG. 7. FIG. 7 is a conceptual diagram illustrating an exemplary CNN-based filter 700 with 14 layers and leaky ReLU as an activation function, according to one or more techniques of this disclosure. In the example of FIG. 7, the input to the CNN-based filter 700 includes multiple sample blocks 702 of a reconstructed picture. In some examples, each of the sample blocks 702 may be a square sample block having a size of N×N. However, this disclosure is not limited to square sample blocks and may generally apply to rectangular sample blocks having a size of N×M. In the example of FIG. 7, a chroma block has half the number of samples of a luma block in the vertical direction and half the number of samples of a luma block in the horizontal direction. Thus, for each N×N chroma block, the input to the CNN-based filter 700 may include four N×N luma blocks. In the example of FIG. 7, the luma blocks are shaded and the chroma blocks are white. Thus, the sample block 702 received as input to the filter 700 can be represented as Y×4+U+V, where Y corresponds to luma, U corresponds to the first chroma component, and V corresponds to the second chroma component.

[0101] Further, the CNN-based filter 700 includes an input layer 704, hidden layers 706A-706N (collectively "hidden layers 706"), and an output layer 708. Specifically, in the example of FIG. 7, the CNN-based filter 700 includes 12 hidden layers for a total of 14 layers. In other examples, the CNN-based filter 700 may have a greater or lesser amount of layers than 14. The input layer 704, hidden layer 706, and output layer 708 each include one of convolutional layers 710A-710N (collectively "convolutional layers 710"). Further, the input layer 704 and hidden layer 706 each include one of LeakyReLU activation layers 712A-712M (collectively "LeakyReLU activation layers 712"). Each of the LeakyReLU activation layers 712 takes the output of a convolutional layer as input and applies a LeakyReLU activation function to the input.

[0102] The output layer 708 of the CNN-based filter 700 generates an intermediate output 701. The intermediate output 701 may include residual samples. These residual samples may differ from the residual samples generated by the residual generation unit 332. The CNN-based filter 700 generates a filtered block 714 by adding the residual samples of the intermediate output 701 to corresponding samples of the sample block 702. The filtered block 714 may include two chroma blocks of size N×N and four luma blocks of size N×N.

[0103] In one example, the optimal value of Alpha is achieved by statistics of the trainable values ​​of PReLU, and the same Alpha value is used for all layers. For example, the video encoder 200 and the video decoder 300 may use the same predetermined value of Alpha in all of the LeakyReLU activation layers 712 of the CNN-based filter 700. In this example, it may not be necessary for the video encoder 200 to signal the value of Alpha (a value of Alpha) in the bitstream. In some examples, to obtain the optimal value of Alpha based on statistics of the values ​​of the trainable values ​​(e.g., a) of the PReLU, a system (e.g., by the video encoder 200, the video decoder 300, or another device or system) may train a network using a PReLU activation function until training converges. The system may then use statistics on the trainable values ​​(e.g., a) of the PReLU activation function of each layer. For example, the system may determine the optimal value of Alpha as the average of the trainable values ​​of the PReLU activation function of the layer.

[0104] As another example, the optimal value of Alpha may be achieved by statistics of the values ​​of the trainable values ​​of PReLU for each layer, and each layer may use its own fixed Alpha value. For example, for each LeakyReLU activation layer 712, the video encoder 200 and the video decoder 300 may use the same predetermined layer-specific value of Alpha in the LeakyReLU activation layer, but different predetermined values ​​of Alpha may be used in different LeakyReLU activation layers 712. In this example, it may not be necessary for the video encoder 200 to signal the value of Alpha in the bitstream. In some examples, to obtain the optimal value of Alpha based on statistics of the values ​​of the trainable values ​​(e.g., a) of PReLU for each layer, a system (e.g., by the video encoder 200, the video decoder 300, or another device or system) may train a network using a PReLU activation function until training converges. The system may then use statistics regarding the trainable values ​​(e.g., a) of the PReLU activation function for each layer. For example, the system may determine the optimal value of Alpha for a layer as equal to the trainable value of the PReLU activation function of the corresponding layer. In another example, the system may determine the optimal value of Alpha for a layer as equal to the average of two or more of the trainable values ​​of two or more PReLU activation functions of the layer.

[0105]

[0103] As another example, the Alpha value in an NN-based filter (e.g., CNN-based filter 700) is a parameter that may be specified when applying the NN-based filter to a picture. For example, in one possible use case involving video coding, on the encoder side, video encoder 200 makes a decision regarding what Alpha value should be used. For example, video encoder 200 may apply an NN-based filter using multiple different predetermined Alpha values ​​and select which one of the Alpha values ​​produces the best result. Video encoder 200 may determine which Alpha value produces the best result by comparing (e.g., by absolute difference sum, etc.) reconstructed blocks or pictures generated using different Alpha values. Furthermore, video encoder 200 may encode information needed to derive the Alpha value as a syntax element in the bitstream. Video decoder 300 reconstructs the Alpha value selected by video encoder 200 and performs the filtering process. Examples of methods for signaling the Alpha value include signaling the Alpha value in the bitstream or signaling the index of a selected Alpha value within a predefined set. A typical value for the Alpha value may be 1 / 4, although other values ​​may be possible.

[0106] 8 is a block diagram illustrating an example video encoder 200 that may implement the techniques of this disclosure. FIG. 8 is provided for illustrative purposes and should not be considered limiting of the techniques broadly illustrated and described in this disclosure. For illustrative purposes, this disclosure describes video encoder 200 in accordance with VVC (ITU-T H.266) and HEVC (ITU-T H.265) techniques. However, the techniques of this disclosure may be implemented by video encoding devices configured for other video coding standards.

[0107] 8, the video encoder 200 includes a video data memory 830, a mode selection unit 802, a residual generation unit 804, a transform processing unit 806, a quantization unit 808, an inverse quantization unit 810, an inverse transform processing unit 812, a reconstruction unit 814, a filter unit 816, a decoded picture buffer (DPB) 818, and an entropy coding unit 820. Any or all of the video data memory 830, the mode selection unit 802, the residual generation unit 804, the transform processing unit 806, the quantization unit 808, the inverse quantization unit 810, the inverse transform processing unit 812, the reconstruction unit 814, the filter unit 816, the DPB 818, and the entropy coding unit 820 may be implemented in one or more processors or processing circuits. For example, the units of the video encoder 200 may be implemented as one or more circuits or logic elements, as part of a hardware circuit, or as part of a processor, an ASIC, or an FPGA. Moreover, video encoder 200 may include additional or alternative processors or processing circuitry for performing these and other functions.

[0108] Video data memory 830 may store video data to be encoded by components of video encoder 200. Video encoder 200 may receive video data stored in video data memory 830, for example, from video source 104 (FIG. 1). DPB 818 may serve as a reference picture memory that stores reference video data for use in predicting subsequently coded video data by video encoder 200. Video data memory 830 and DPB 818 may be formed by any of a variety of memory devices, such as synchronous dynamic random access memory (DRAM), including DRAM (SDRAM), magnetoresistive RAM (MRAM), resistive RAM (RRAM), or other types of memory devices. Video data memory 830 and DPB 818 may be provided by the same memory device or separate memory devices. In various examples, video data memory 830 may be on-chip with other components of video encoder 200, as shown, or off-chip relative to those components.

[0109] In this disclosure, references to video data memory 830 should not be construed as limited to memory internal to video encoder 200 unless specifically so described, nor should they be construed as limited to memory external to video encoder 200 unless specifically so described. Instead, references to video data memory 830 should be understood as a reference memory that stores video data that video encoder 200 receives for encoding (e.g., video data for a current block to be encoded). Memory 106 of FIG. 1 may also provide temporary storage of outputs from various units of video encoder 200.

[0110] The various units in FIG. 8 are shown to aid in understanding the operations performed by video encoder 200. The units may be implemented as fixed-function circuits, programmable circuits, or a combination thereof. A fixed-function circuit refers to a circuit that provides a specific function and is preset with respect to the operations that may be performed. A programmable circuit refers to a circuit that can be programmed to perform various tasks and to provide flexible functionality in the operations that may be performed. For example, a programmable circuit may execute software or firmware that causes the programmable circuit to operate in a manner defined by the software or firmware instructions. A fixed-function circuit may execute software instructions (e.g., to receive or output parameters), but the types of operations that the fixed-function circuit performs are generally invariant. In some examples, one or more of the units may be separate circuit blocks (fixed function or programmable), and in some examples, one or more of the units may be integrated circuits.

[0111]

[0109] Video encoder 200 may include an arithmetic logic unit (ALU), a basic functional unit (EFU), a programmable core formed from digital circuits, analog circuits, and / or programmable circuits. In examples in which the operations of video encoder 200 are implemented using software executed by programmable circuits, memory 106 (FIG. 1) may store instructions (e.g., object code) of the software that video encoder 200 receives and executes, or another memory (not shown) within video encoder 200 may store such instructions.

[0112]

[0110] The video data memory 830 is configured to store received video data. The video encoder 200 may retrieve pictures of the video data from the video data memory 830 and provide the video data to the residual generation unit 804 and the mode selection unit 802. The video data in the video data memory 830 may be raw video data to be encoded.

[0113] The mode select unit 802 includes a motion estimation unit 822, a motion compensation unit 824, and an intra prediction unit 826. The mode select unit 802 may include additional functional units for performing video prediction according to other prediction modes. By way of example, the mode select unit 802 may include a palette unit, an intra block copy unit (which may be part of the motion estimation unit 822 and / or the motion compensation unit 824), an affine unit, a linear model (LM) unit, etc.

[0114] The mode selection unit 802 generally coordinates multiple encoding passes to test combinations of encoding parameters and obtained rate-distortion values ​​for such combinations. The encoding parameters may include partitioning of CTUs into CUs, prediction modes for CUs, transform types for residual data of CUs, quantization parameters for residual data of CUs, etc. The mode selection unit 802 may finally select a combination of encoding parameters that has a rate-distortion value that is better than other tested combinations.

[0115] Video encoder 200 may partition a picture retrieved from video data memory 830 into a series of CTUs, encapsulating one or more CTUs in a slice. Mode select unit 802 may partition the CTUs of the picture according to a tree structure, such as the QTBT structure or quadtree structure of HEVC described above. As described above, video encoder 200 may form one or more CUs from partitioning the CTUs according to the tree structure. Such CUs may also be generally referred to as “video blocks” or “blocks.”

[0116] Generally, the mode selection unit 802 also controls its components (e.g., the motion estimation unit 822, the motion compensation unit 824, and the intra prediction unit 826) to generate a prediction block for a current block (e.g., the current CU, or in HEVC, the overlapping portion of the PU and TU). For inter prediction of the current block, the motion estimation unit 822 may perform motion search to identify one or more closely matching reference blocks in one or more reference pictures (e.g., one or more previously coded pictures stored in the DPB 818). In particular, the motion estimation unit 822 may calculate a value representing how similar a potential reference block is to the current block according to, for example, the sum of absolute differences (SAD), the sum of squared differences (SSD), the mean absolute difference (MAD), the mean squared difference (MSD), etc. The motion estimation unit 822 may generally perform these calculations using sample-by-sample differences between the current block and the reference block under consideration. Motion estimation unit 822 may identify the reference block with the lowest value resulting from these calculations, indicating the reference block that most closely matches the current block.

[0117] The motion estimation unit 822 may form one or more motion vectors (MVs) that define the position of a reference block in a reference picture relative to the position of a current block in the current picture. The motion estimation unit 822 may then provide the motion vectors to the motion compensation unit 824. For example, in unidirectional inter prediction, the motion estimation unit 822 may provide a single motion vector, while in bidirectional inter prediction, the motion estimation unit 822 may provide two motion vectors. The motion compensation unit 824 may then generate a predictive block using the motion vectors. For example, the motion compensation unit 824 may use the motion vectors to retrieve data of the reference block. As another example, if the motion vectors have sub-sample precision, the motion compensation unit 824 may interpolate values ​​for the predictive block according to one or more interpolation filters. Moreover, in bidirectional inter prediction, the motion compensation unit 824 may retrieve data for the two reference blocks identified by the respective motion vectors and combine the retrieved data, for example, through sample-wise averaging or weighted averaging.

[0118] As another example, for intra prediction, or intra predictive coding, the intra prediction unit 826 may generate a predictive block from samples neighboring a current block. For example, in a directional mode, the intra prediction unit 826 may generally mathematically combine values ​​of neighboring samples and populate these calculated values ​​in a defined direction across the current block to create a predictive block. As another example, in a DC mode, the intra prediction unit 826 may calculate an average of neighboring samples for the current block and generate a predictive block to include this resulting average for each sample of the predictive block.

[0119] The mode select unit 802 provides the prediction block to the residual generation unit 804. The residual generation unit 804 receives a raw, uncoded version of the current block from the video data memory 830 and receives the prediction block from the mode select unit 802. The residual generation unit 804 calculates sample-by-sample differences between the current block and the prediction block. The resulting sample-by-sample differences define a residual block for the current block. In some examples, the residual generation unit 804 may also determine differences between sample values ​​in the residual block to generate the residual block using residual differential pulse code modulation (RDPCM). In some examples, the residual generation unit 804 may be formed using one or more subtractor circuits that perform binary subtraction.

[0120] In examples in which the mode select unit 802 partitions CUs into PUs, each PU may be associated with a luma prediction unit and a corresponding chroma prediction unit. The video encoder 200 and the video decoder 300 may support PUs having various sizes. As indicated above, the size of a CU may refer to the size of the luma coding block of the CU, and the size of a PU may refer to the size of the luma prediction unit of the PU. Assuming that the size of a particular CU is 2N×2N, the video encoder 200 may support a PU size of 2N×2N or N×N for intra prediction and a symmetric PU size of 2N×2N, 2N×N, N×2N, N×N, or the like for inter prediction. The video encoder 200 and the video decoder 300 may also support asymmetric partitioning for PU sizes of 2N×nU, 2N×nD, nL×2N, and nR×2N for inter prediction.

[0121] In examples where the mode select unit 802 does not further partition CUs into PUs, each CU may be associated with a luma coding block and a corresponding chroma coding block. As described above, the size of a CU may refer to the size of the luma coding block of the CU. The video encoder 200 and the video decoder 300 may support CU sizes of 2N×2N, 2N×N, or N×2N.

[0122] In other video coding techniques, such as intra block copy mode coding, affine mode coding, and linear model (LM) mode coding, for example, the mode select unit 802 generates a predictive block for the current block being coded via a respective unit associated with the coding technique. In some examples, such as palette mode coding, the mode select unit 802 may not generate a predictive block, but instead generate syntax elements that indicate how the block should be reconstructed based on a selected palette. In such modes, the mode select unit 802 may provide these syntax elements to be coded to the entropy coding unit 820.

[0123]

[0121] As described above, the residual generation unit 804 receives video data for a current block and a corresponding predictive block. The residual generation unit 804 then generates a residual block for the current block. To generate the residual block, the residual generation unit 804 calculates sample-by-sample differences between the predictive block and the current block.

[0124] Transform processing unit 806 applies one or more transforms to the residual block to generate a block of transform coefficients (referred to herein as a "transform coefficient block"). Transform processing unit 806 may apply various transforms to the residual block to form the transform coefficient block. For example, transform processing unit 806 may apply a discrete cosine transform (DCT), a directional transform, a Karhunen-Loeve transform (KLT), or a conceptually similar transform to the residual block. In some examples, transform processing unit 806 may perform multiple transforms on the residual block, e.g., a linear transform and a quadratic transform, such as a rotation transform. In some examples, transform processing unit 806 does not apply a transform to the residual block.

[0125] The quantization unit 808 may quantize the transform coefficients in the transform coefficient block to produce a quantized transform coefficient block. The quantization unit 808 may quantize the transform coefficients of the transform coefficient block according to a quantization parameter (QP) value associated with the current block. The video encoder 200 (e.g., via the mode selection unit 802) may adjust the degree of quantization applied to the transform coefficient block associated with the current block by adjusting the QP value associated with the CU. Quantization may result in a loss of information, and therefore, the quantized transform coefficients may have lower precision than the original transform coefficients produced by the transform processing unit 806.

[0126] The inverse quantization unit 810 and the inverse transform processing unit 812 may apply inverse quantization and inverse transform to the quantized transform coefficient block, respectively, to reconstruct a residual block from the transform coefficient block. The reconstruction unit 814 may produce a reconstructed block that corresponds to the current block (potentially with some distortion) based on the reconstructed residual block and the predictive block generated by the mode select unit 802. For example, the reconstruction unit 814 may add samples of the reconstructed residual block to corresponding samples from the predictive block generated by the mode select unit 802 to produce the reconstructed block.

[0127] Filter unit 816 may perform one or more filter operations on the reconstructed blocks. For example, filter unit 816 may perform a deblocking operation to reduce blockiness artifacts along the edges of a CU. The operations of filter unit 816 may be skipped in some examples. In some examples, filter unit 816 may apply a CNN-based filter that uses a LeakyReLU activation function, such as CNN-based filter 700. Thus, in some examples, reconstruction unit 814 may reconstruct blocks of a picture, where a block may be a square or rectangular area of ​​a picture and may or may not correspond to a coding block, a prediction block, etc. Filter unit 816 may apply CNN-based filter 700 to the block, where CNN-based filter 700 uses a LeakyReLU activation function.

[0128] The video encoder 200 stores the reconstructed blocks in the DPB 818. For example, in examples where the operations of the filter unit 816 are not performed, the reconstruction unit 814 may store the reconstructed blocks in the DPB 818. In examples where the operations of the filter unit 816 are performed, the filter unit 816 may store the filtered reconstructed blocks in the DPB 818. The motion estimation unit 822 and the motion compensation unit 824 may retrieve reference pictures formed from the reconstructed (and potentially filtered) blocks from the DPB 818 to inter-predict blocks of a later-encoded picture. Furthermore, the intra-prediction unit 826 may use the reconstructed blocks in the DPB 818 of the current picture to intra-predict other blocks in the current picture.

[0129] Generally, the entropy coding unit 820 may entropy code syntax elements received from other functional components of the video encoder 200. For example, the entropy coding unit 820 may entropy code quantized transform coefficient blocks from the quantization unit 808. As another example, the entropy coding unit 820 may entropy code predictive syntax elements (e.g., motion information for inter prediction or intra-mode information for intra prediction) from the mode select unit 802. The entropy coding unit 820 may perform one or more entropy coding operations on syntax elements, which are another example of video data, to generate entropy-coded data. For example, the entropy encoding unit 820 may perform a context-adaptive variable length coding (CAVLC) operation, a CABAC operation, a variable-to-variable (V2V) length coding operation, a syntax-based context-adaptive binary arithmetic coding (SBAC) operation, a probability interval partitioned entropy (PIPE) coding operation, an exponential-Golomb coding operation, or another type of entropy coding operation on the data. In some examples, the entropy encoding unit 820 may operate in a bypass mode in which syntax elements are not entropy coded.

[0130] Video encoder 200 may output a bitstream that includes entropy-encoded syntax elements needed to reconstruct blocks of a slice or picture. For example, entropy encoding unit 820 may output a bitstream.

[0131]

[0129] The operations described above are described with respect to blocks. Such descriptions may be understood as being operations for luma coding blocks and / or chroma coding blocks. As described above, in some examples, the luma coding blocks and chroma coding blocks are luma and chroma components of a CU. In some examples, the luma coding blocks and chroma coding blocks are luma and chroma components of a PU.

[0132]

[0130] In some examples, operations performed with respect to luma coding blocks do not need to be repeated for chroma coding blocks. As an example, operations for identifying motion vectors (MVs) and reference pictures for luma coding blocks do not need to be repeated to identify MVs and reference pictures for chroma blocks. Rather, MVs for luma coding blocks may be scaled to determine MVs for chroma blocks, and the reference pictures may be the same. As another example, the intra prediction process may be the same for luma coding blocks and chroma coding blocks.

[0133]

[0131] Video encoder 200 represents an example of a device configured to encode video data, including a memory configured to store the video data and one or more processing units implemented in a circuit, the one or more processing units configured to reconstruct blocks of the video data and apply a CNN-based filter to the blocks, wherein the CNN-based filter uses a LeakyReLU activation function.

[0134] 9 is a block diagram illustrating an example video decoder 300 that may implement the techniques of this disclosure. FIG. 9 is provided for illustrative purposes and is not intended to limit the techniques broadly illustrated and described in this disclosure. For illustrative purposes, this disclosure describes the video decoder 300 in accordance with VVC (ITU-T H.266 under development) and HEVC (ITU-T H.265) techniques. However, the techniques of this disclosure may be implemented by video coding devices configured for other video coding standards.

[0135] 9, the video decoder 300 includes a coded picture buffer (CPB) memory 920, an entropy decoding unit 902, a prediction processing unit 904, an inverse quantization unit 906, an inverse transform processing unit 908, a reconstruction unit 910, a filter unit 912, and a decoded picture buffer (DPB) 914. Any or all of the CPB memory 920, the entropy decoding unit 902, the prediction processing unit 904, the inverse quantization unit 906, the inverse transform processing unit 908, the reconstruction unit 910, the filter unit 912, and the DPB 914 may be implemented in one or more processors or processing circuits. For example, the units of the video decoder 300 may be implemented as one or more circuits or logic elements, as part of a hardware circuit, or as part of a processor, ASIC, or FPGA. Moreover, the video decoder 300 may include additional or alternative processors or processing circuits for performing these and other functions.

[0136] The prediction processing unit 904 includes a motion compensation unit 916 and an intra prediction unit 918. The prediction processing unit 904 may include additional units for performing prediction according to other prediction modes. By way of example, the prediction processing unit 904 may include a palette unit, an intra block copy unit (which may form part of the motion compensation unit 916), 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] The CPB memory 920 may store video data, such as an encoded video bitstream, to be decoded by components of the video decoder 300. The video data stored in the CPB memory 920 may be retrieved, for example, from the computer-readable medium 110 (FIG. 1). The CPB memory 920 may include a CPB that stores encoded video data (e.g., syntax elements) from the encoded video bitstream. The CPB memory 920 may also store video data other than syntax elements of coded pictures, such as temporary data representing output from various units of the video decoder 300. The DPB 914 generally stores decoded pictures that the video decoder 300 may output and / or use as reference video data when decoding subsequent data or pictures of the encoded video bitstream. The CPB memory 920 and the DPB 914 may be formed by any of a variety of memory devices, such as DRAM, including SDRAM, MRAM, RRAM, or other types of memory devices. The CPB memory 920 and the DPB 914 may be provided by the same memory device or separate memory devices. In various examples, the CPB memory 920 may be on-chip with other components of the video decoder 300 or off-chip relative to those components.

[0138] Additionally or alternatively, in some examples, video decoder 300 may retrieve coded video data from memory 120 (FIG. 1). That is, memory 120 may store data as described above in conjunction with CPB memory 920. Similarly, memory 120 may store instructions to be executed by video decoder 300 when some or all of the functionality of video decoder 300 is implemented in software to be executed by processing circuitry of video decoder 300.

[0139] The various units shown in FIG. 9 are presented to aid in understanding the operations performed by video decoder 300. The units may be implemented as fixed-function circuits, programmable circuits, or a combination thereof. As with FIG. 8, a fixed-function circuit refers to a circuit that provides a specific function and is preset with respect to the operations that may be performed. A programmable circuit refers to a circuit that may be programmed to perform various tasks and to provide flexible functionality in the operations that may be performed. For example, a programmable circuit may execute software or firmware that causes the programmable circuit to operate in a manner defined by the software or firmware instructions. A fixed-function circuit may execute software instructions (e.g., to receive parameters or output parameters), but the types of operations that the fixed-function circuit performs are generally invariant. In some examples, one or more of the units may be separate circuit blocks (fixed function or programmable), and in some examples, one or more of the units may be an integrated circuit.

[0140]

[0138] The video decoder 300 may include a programmable core formed from an ALU, an EFU, digital circuits, analog circuits, and / or programmable circuits. In examples in which the operations of the video decoder 300 are performed by software executing on programmable circuits, on-chip or off-chip memory may store instructions (e.g., object code) of the software that the video decoder 300 receives and executes.

[0141] The entropy decoding unit 902 may receive the encoded video data from the CPB and entropy decode the video data to recover the syntax elements. The prediction processing unit 904, the inverse quantization unit 906, the inverse transform processing unit 908, the reconstruction unit 910, and the filter unit 912 may generate decoded video data based on the syntax elements extracted from the bitstream. In some examples, the filter unit 912 may apply a CNN-based filter that uses a LeakyReLU activation function.

[0142] Generally, the video decoder 300 reconstructs a picture on a block-by-block basis. The video decoder 300 may perform a reconstruction operation on each block individually (where the block currently being reconstructed, i.e., decoded, may be referred to as the “current block”).

[0143] The entropy decoding unit 902 may entropy decode syntax elements defining quantized transform coefficients of a quantized transform coefficient block, as well as transform information such as a quantization parameter (QP) and / or one or more transform mode indications. The inverse quantization unit 906 may use the QP associated with the quantized transform coefficient block to determine the degree of quantization and, similarly, the degree of inverse quantization to apply. The inverse quantization unit 906 may perform, for example, a bitwise left shift operation to inverse quantize the quantized transform coefficients. The inverse quantization unit 906 may thereby form a transform coefficient block including the transform coefficients.

[0144] After the inverse quantization unit 906 forms the transform coefficient blocks, the inverse transform processing unit 908 may apply one or more inverse transforms to the transform coefficient blocks to generate residual blocks associated with the current block. For example, the inverse transform processing unit 908 may apply an inverse DCT, an inverse integer transform, an inverse Karhunen-Loeve transform (KLT), an inverse rotational transform, an inverse transform, or another inverse transform to the transform coefficient blocks.

[0145] Further, prediction processing unit 904 generates a predictive block according to the prediction information syntax element entropy decoded by entropy decoding unit 902. For example, if the prediction information syntax element indicates that the current block is inter-predicted, motion compensation unit 916 may generate a predictive block. In this case, the prediction information syntax element may indicate a reference picture in DPB 914 from which to retrieve a reference block, as well as a motion vector that identifies the location of the reference block in the reference picture relative to the location of the current block in the current picture. Motion compensation unit 916 may generally perform the inter-prediction process in a manner substantially similar to that described with respect to motion compensation unit 824 (FIG. 8).

[0146] As another example, if the prediction information syntax element indicates that the current block is intra-predicted, the intra prediction unit 918 may generate a predictive block according to the intra-prediction mode indicated by the prediction information syntax element. Again, the intra prediction unit 918 may generally perform the intra prediction process in a manner substantially similar to that described with respect to the intra prediction unit 826 (FIG. 8). The intra prediction unit 918 may retrieve data of neighboring samples for the current block from the DPB 914.

[0147]

[0145] The reconstruction unit 910 may reconstruct the current block using the predictive block and the residual block. For example, the reconstruction unit 910 may add samples of the residual block to corresponding samples of the predictive block to reconstruct the current block.

[0148] The filter unit 912 may perform one or more filter operations on the reconstructed block. For example, the filter unit 912 may perform a deblocking operation to reduce blockiness artifacts along the edges of the reconstructed block. The operations of the filter unit 912 are not necessarily performed in all examples. In some examples, the filter unit 912 may apply a CNN-based filter that uses a LeakyReLU activation function, such as the CNN-based filter 700. Thus, in some examples, the reconstruction unit 910 may reconstruct a block of a picture, where the block may be a square or rectangular area of ​​the picture and may or may not correspond to a coding block, a prediction block, etc. The filter unit 912 may apply the CNN-based filter 700 to the block, where the CNN-based filter 700 uses a LeakyReLU activation function.

[0149] The video decoder 300 may store the reconstructed block in the DPB 914. For example, in examples where the operations of the filter unit 912 are not performed, the reconstruction unit 910 may store the reconstructed block in the DPB 914. In examples where the operations of the filter unit 912 are performed, the filter unit 912 may store the filtered reconstructed block in the DPB 914. As described above, the DPB 914 may provide reference information to the prediction processing unit 904, such as samples of the current picture for intra prediction and previously decoded pictures for subsequent motion compensation. Moreover, the video decoder 300 may output the decoded picture (e.g., the decoded video) from the DPB 914 for subsequent presentation on a display device, such as the display device 118 of FIG. 1 .

[0150]

[0148] Thus, the video decoder 300 represents an example of a video decoding device that includes a memory configured to store video data and one or more processing units implemented in a circuit, the one or more processing units configured to reconstruct blocks of the video data and apply a CNN-based filter to the blocks, wherein the CNN-based filter uses a LeakyReLU activation function.

[0151] 10 is a flowchart illustrating an example method for encoding a current block in accordance with the techniques of this disclosure. The current block may comprise a current CU. Although described with respect to video encoder 200 (FIGS. 1 and 8), it should be understood that other devices may be configured to implement a method similar to that of FIG.

[0152] In this example, the video encoder 200 first predicts the current block (350). For example, the video encoder 200 may form a predictive block for the current block. The video encoder 200 may then calculate a residual block for the current block (352). To calculate the residual block, the video encoder 200 may calculate the difference between the original uncoded block and the predictive block for the current block. The video encoder 200 may then transform the residual block and quantize the transform coefficients of the residual block (354). The video encoder 200 may then scan the quantized transform coefficients of the residual block (356). During or following the scan, the video encoder 200 may entropy code the transform coefficients (358). For example, the video encoder 200 may code the transform coefficients using CAVLC or CABAC. The video encoder 200 may then output entropy-coded data for the block (360).

[0153] 11 is a flowchart illustrating an example method for decoding a current block of video data in accordance with the techniques of this disclosure. The current block may comprise a current CU. Although described with respect to video decoder 300 (FIGS. 1 and 9), it should be understood that other devices may be configured to implement a method similar to that of FIG.

[0154] The video decoder 300 may receive entropy-coded data for the current block, such as entropy-coded prediction information and entropy-coded data for the transform coefficients of a residual block corresponding to the current block (370). The video decoder 300 may entropy decode the entropy-coded data to determine prediction information for the current block and to reconstruct the transform coefficients of the residual block (372). The video decoder 300 may predict the current block, e.g., using the intra-prediction or inter-prediction mode indicated by the prediction information for the current block, to calculate a predictive block for the current block (374). The video decoder 300 may then inverse scan the reconstructed transform coefficients to create a block of quantized transform coefficients (376). The video decoder 300 may then dequantize the transform coefficients and apply an inverse transform to the transform coefficients to produce a residual block (378). The video decoder 300 may finally decode the current block by combining the predictive block and the residual block (380). In some examples, the video decoder 300 may further apply a CNN-based filter to the decoded block, where the CNN-based filter uses a LeakyReLU activation function.

[0155] FIG. 12 is a flowchart illustrating an example operation of a video coder (e.g., video encoder 200 or video decoder 300) in accordance with one or more techniques of this disclosure. In the example of FIG. 12, the video coder reconstructs a block of video data (1200). The block of video data may be a rectangular array of samples. The block may or may not correspond to a transform block, a predictive block, a coding block, a coding tree block, or a block of a type used in the coding process. For example, the video coder (e.g., reconstruction unit 814 (FIG. 8) or reconstruction unit 910 (FIG. 9)) may reconstruct a picture including the block by adding residual data to the corresponding predictive block.

[0156]

[0154] Furthermore, the video coder applies (1202) a CNN-based filter (e.g., CNN-based filter 700) to the reconstructed block. The CNN-based filter uses a LeakyReLU activation function. As described above, the LeakyReLU activation function is

[0157]

number

[0158] where y is the output value of a convolutional layer of the CNN and Alpha is a fixed parameter. In some examples, the CNN-based filter includes multiple convolutional layers, and the value of Alpha is the same for each of the convolutional layers. In some examples, the CNN-based filter includes multiple convolutional layers, and the value of Alpha is different for two or more of the convolutional layers.

[0159] In some examples where the video coder is the video decoder 300, the video decoder 300 may determine the value of Alpha based on one or more syntax elements signaled in a bitstream comprising an encoded representation of the video data. For example, the value of Alpha may be signaled in the bitstream. In other words, a syntax element may directly specify the value of Alpha. In another example, an index of Alpha in a predefined set is signaled in the bitstream. In this example, the video decoder 300 may determine the value of Alpha in the predefined set to which the index of Alpha corresponds.

[0160] In some examples where the video coder is video encoder 200, the CNN-based filter includes multiple convolutional layers (e.g., convolutional layer 710), and video encoder 200 may determine a single value of Alpha for use in a LeakyReLU activation function applied to output values ​​of the convolutional layers. Video encoder 200 may encode, in a bitstream comprising an encoded representation of the video data, one or more syntax elements that provide information needed by a video decoder to derive a value of Alpha. In some examples, video encoder 200 signals, in the bitstream, an index of Alpha in a predefined set. In this example, video encoder 200 may determine a value of Alpha in the predefined set to which the index of Alpha corresponds. In other examples, video encoder 200 may determine multiple values ​​of Alpha for use in a LeakyReLU activation function applied to output values ​​of two or more different convolutional layers of the multiple convolutional layers (e.g., convolutional layer 710). In such an example, video encoder 200 may encode, in a bitstream comprising an encoded representation of the video data, one or more syntax elements that provide information needed by a video decoder to derive the value of Alpha.

[0161]

[0157] The following is a non-limiting list of aspects that may be in accordance with one or more aspects of the present disclosure.

[0162]

[0158] Aspect 1A. A method for coding video data, the method comprising: reconstructing a block of the video data; and applying a convolutional neural network (CNN)-based filter to the block, wherein the CNN-based filter uses a LeakyReLU activation function.

[0163] Aspect 2A. The LeakyReLU activation function is:

[0164]

number

[0165] where y is an output value of a convolutional layer of the CNN, and Alpha is a fixed parameter.

[0166]

[0160] Aspect 3A. The method of aspect 2A, wherein the CNN includes multiple convolutional layers and the value of Alpha is the same for each of the convolutional layers.

[0167]

[0161] Aspect 4A. The method of aspect 2A, wherein the CNN includes multiple convolutional layers and the value of Alpha is different for two or more of the convolutional layers.

[0168]

[0162] Aspect 5A. A method according to any of aspects 2A to 4A, further comprising determining the value of Alpha based on one or more syntax elements signaled in a bitstream comprising an encoded representation of the video data.

[0169] Aspect 6A. The method of aspect 5A, wherein the value of Alpha is signaled in the bitstream.

[0170] Aspect 7A. The method of aspect 5A, wherein the index of Alpha within a predefined set is signaled in the bitstream.

[0171] Aspect 8A. The method of any of aspects 1A to 7A, wherein the coding comprises decoding.

[0172] Aspect 9A. The method of any of aspects 1A to 7A, wherein coding comprises encoding.

[0173] Aspect 10A. A device for coding video data, the device comprising one or more means for performing the method of any of aspects 1A to 9A.

[0174] Aspect 11A. The device of Aspect 10A, wherein the one or more means comprise one or more processors implemented in circuitry.

[0175] Aspect 12A. The device of any of aspects 10A and 11A, further comprising a memory for storing video data.

[0176] Aspect 13A. The device of any of aspects 10A to 12A, further comprising a display configured to display the decoded video data.

[0177] Aspect 14A. The device of any of aspects 10A to 13A, wherein the device comprises one or more of a camera, a computer, a mobile device, a broadcast receiver device, or a set-top box.

[0178] Aspect 15A. The device of any of aspects 10A to 14A, wherein the device comprises a video decoder.

[0179] Aspect 16A. The device of any of aspects 10A to 15A, wherein the device comprises a video encoder.

[0180] Aspect 17A. A computer-readable storage medium storing instructions that, when executed, cause one or more processors to perform a method according to any of aspects 1A to 9A.

[0181]

[0175] Aspect 1B: A method for encoding or decoding video data includes reconstructing a block of the video data and applying a convolutional neural network (CNN)-based filter to the reconstructed block, wherein the CNN-based filter uses a LeakyReLU activation function.

[0182] Aspect 2B: The LeakyReLU activation function is

[0183]

number

[0184] where y is an output value of a convolutional layer of the CNN of the CNN-based filter, and Alpha is a fixed parameter.

[0185]

[0177] Aspect 3B: The method of aspect 2B, wherein the CNN includes multiple convolutional layers and the value of Alpha is the same for each of the convolutional layers.

[0186]

[0178] Aspect 4B: The method of aspect 2B, wherein the CNN includes multiple convolutional layers and the value of Alpha is different for two or more of the convolutional layers.

[0187]

[0179] Aspect 5B: The method of aspect 2B, further comprising determining a value for Alpha based on one or more syntax elements signaled in a bitstream comprising the encoded representation of the video data.

[0188] Aspect 6B: The method of aspect 5B, wherein the value of Alpha is signaled in the bitstream.

[0189]

[0181] Aspect 7B: A method as described in aspect 5B, wherein an index of Alpha within a predefined set is signaled in the bitstream, and wherein determining the value of Alpha comprises determining the value of Alpha in the predefined set to which the index of Alpha corresponds.

[0190]

[0182] Aspect 8B: The method described in aspect 2B, wherein the CNN includes multiple convolutional layers, and the method further comprises determining a single value of Alpha for use in a LeakyReLU activation function applied to output values ​​of the convolutional layers, and encoding, in a bitstream comprising an encoded representation of the video data, one or more syntax elements that provide information required by a video decoder to derive the value of Alpha.

[0191] Aspect 9B. The method of aspect 8B, wherein one or more syntax elements directly specify a value for Alpha.

[0192] Aspect 10B. The method of aspect 8B, wherein one or more syntax elements specify an index of Alpha within a predefined set.

[0193]

[0185] Aspect 11B: The method described in aspect 2B, wherein the CNN includes multiple convolutional layers, and the method further comprises determining multiple values ​​of Alpha for use in a LeakyReLU activation function applied to output values ​​of two or more different convolutional layers of the multiple convolutional layers, and encoding, in a bitstream comprising an encoded representation of the video data, one or more syntax elements that provide information required by a video decoder to derive the values ​​of Alpha.

[0194] Aspect 12B. The method of aspect 11B, wherein one or more syntax elements directly specify a value for Alpha.

[0195] Aspect 13B. The method of aspect 11B, wherein one or more syntax elements specify an index of Alpha within a predefined set.

[0196]

[0188] Aspect 14B: A method as described in aspect 1B, wherein the method comprises decoding video data, and the decoding of the video data includes reconstructing blocks of the video data and applying a CNN-based filter to the reconstructed blocks.

[0197]

[0189] Aspect 15B: A method as described in aspect 1B, wherein the method comprises encoding video data, and the encoding of the video data includes reconstructing blocks of the video data and applying a CNN-based filter to the reconstructed blocks.

[0198]

[0190] Aspect 16B: A device for encoding or decoding video data includes a memory for storing the video data and one or more processors implemented in a circuit, the one or more processors configured to reconstruct blocks of the video data and apply a convolutional neural network (CNN)-based filter to the reconstructed blocks, wherein the CNN-based filter uses a LeakyReLU activation function.

[0199] Aspect 17B: The LeakyReLU activation function is:

[0200]

number

[0201] where y is an output value of a convolutional layer of the CNN of the CNN-based filter, and Alpha is a fixed parameter.

[0202]

[0192] Aspect 18B: A device described in aspect 17B, wherein the CNN includes multiple convolutional layers and the value of Alpha is the same for each of the convolutional layers.

[0203]

[0193] Aspect 19B: A device described in aspect 17B, wherein the CNN includes multiple convolutional layers and the value of Alpha is different for two or more of the convolutional layers.

[0204]

[0194] Aspect 20B: The device described in aspect 17B, wherein the one or more processors are further configured to determine a value for Alpha based on one or more syntax elements signaled in a bitstream comprising an encoded representation of the video data.

[0205]

[0195] Aspect 21B: The device of aspect 20B, wherein the value of Alpha is signaled in the bitstream.

[0206]

[0196] Aspect 22B: A device described in aspect 20B, wherein an index of Alpha within a predefined set is signaled in the bitstream, and wherein determining the value of Alpha comprises determining the value of Alpha in the predefined set to which the index of Alpha corresponds.

[0207]

[0197] Aspect 23B: The device described in Aspect 17B, wherein the CNN-based filter includes multiple convolutional layers, and wherein one or more processors are further configured to determine a single value of Alpha for use in a LeakyReLU activation function applied to the output values ​​of the convolutional layers, and to encode, in a bitstream comprising an encoded representation of the video data, one or more syntax elements that provide information required by a video decoder to derive the value of Alpha.

[0208]

[0198] Aspect 24B: The device of Aspect 17B, wherein the CNN includes multiple convolutional layers, and wherein one or more processors are further configured to determine multiple values ​​of Alpha for use in a LeakyReLU activation function applied to output values ​​of two or more different convolutional layers of the multiple convolutional layers, and to encode, in a bitstream comprising an encoded representation of the video data, one or more syntax elements that provide information required by a video decoder to derive the values ​​of Alpha.

[0209]

[0199] Aspect 25B: The device of aspect 16B, further comprising a display configured to display the decoded video data.

[0210] Aspect 26B: The device of aspect 16B, wherein the device comprises one or more of a camera, a computer, a mobile device, a broadcast receiver device, or a set-top box.

[0211]

[0201] Aspect 27B: The device of aspect 16B, wherein the device comprises a video decoder.

[0212]

[0202] Aspect 28B: The device of aspect 16B, wherein the device comprises a video encoder.

[0213]

[0203] Aspect 29B: A device for coding video data includes means for reconstructing a block of video data and means for applying a convolutional neural network (CNN)-based filter to the reconstructed block, wherein the CNN-based filter uses a LeakyReLU activation function.

[0214]

[0204] Aspect 30B: A computer-readable storage medium having stored thereon instructions that, when executed, cause one or more processors to reconstruct blocks of video data and apply a convolutional neural network (CNN)-based filter to the reconstructed blocks, wherein the CNN-based filter uses a LeakyReLU activation function.

[0215]

[0205] Aspect 1C: A method for encoding or decoding video data includes reconstructing a block of the video data and applying a convolutional neural network (CNN)-based filter to the reconstructed block, wherein the CNN-based filter uses a LeakyReLU activation function.

[0216] Aspect 2C: The LeakyReLU activation function is

[0217]

number

[0218] where y is an output value of a convolutional layer of the CNN of the CNN-based filter, and Alpha is a fixed parameter.

[0219]

[0207] Aspect 3C: The method of aspect 2C, wherein the CNN includes multiple convolutional layers and the value of Alpha is the same for each of the convolutional layers.

[0220]

[0208] Aspect 4C: The method of aspect 2C, wherein the CNN includes multiple convolutional layers and the value of Alpha is different for two or more of the convolutional layers.

[0221]

[0209] Aspect 5C: A method according to any of aspects 2C to 4C, further comprising determining the value of Alpha based on one or more syntax elements signaled in a bitstream comprising an encoded representation of the video data.

[0222]

[0210] Aspect 6C: The method of aspect 5C, wherein the value of Alpha is signaled in the bitstream.

[0223]

[0211] Aspect 7C: A method as described in aspect 5C, wherein an index of Alpha within a predefined set is signaled in the bitstream, and wherein determining the value of Alpha comprises determining the value of Alpha in the predefined set to which the index of Alpha corresponds.

[0224]

[0212] Aspect 8C: A method according to any of aspects 2C to 3C or 5C to 7C, wherein the CNN includes multiple convolutional layers, and the method further comprises determining a single value of Alpha for use in a LeakyReLU activation function applied to output values ​​of the convolutional layers, and encoding, in a bitstream comprising the encoded representation of the video data, one or more syntax elements that provide information required by a video decoder to derive the value of Alpha.

[0225] Aspect 9C. The method of aspect 8C, wherein one or more syntax elements directly specify a value for Alpha.

[0226] Aspect 10C. The method of aspect 8C, wherein one or more syntax elements specify an index of Alpha within a predefined set.

[0227]

[0215] Aspect 11C: A method according to any of aspects 2C or 4C to 7C, wherein the CNN includes multiple convolutional layers, and the method further comprises determining multiple values ​​of Alpha for use in a LeakyReLU activation function applied to output values ​​of two or more different convolutional layers of the multiple convolutional layers, and encoding, in a bitstream comprising the encoded representation of the video data, one or more syntax elements that provide information required by a video decoder to derive the values ​​of Alpha.

[0228] Aspect 12C. The method of aspect 11C, wherein one or more syntax elements directly specify a value for Alpha.

[0229] Aspect 13C. The method of aspect 11C, wherein one or more syntax elements specify an index of Alpha within a predefined set.

[0230]

[0218] Aspect 14C: A method described in any of aspects 1C to 13C, wherein the method comprises decoding video data, and the decoding of the video data includes reconstructing blocks of the video data and applying a CNN-based filter to the reconstructed blocks.

[0231]

[0219] Aspect 15C: A method described in any of aspects 1B to 14C, wherein the method comprises encoding video data, and encoding the video data includes reconstructing blocks of the video data and applying a CNN-based filter to the reconstructed blocks.

[0232]

[0220] Aspect 16C: A device for encoding or decoding video data includes a memory for storing the video data and one or more processors implemented in a circuit, the one or more processors configured to reconstruct blocks of the video data and apply a convolutional neural network (CNN)-based filter to the reconstructed blocks, wherein the CNN-based filter uses a LeakyReLU activation function.

[0233] Aspect 17C: The LeakyReLU activation function is:

[0234]

number

[0235] where y is an output value of a convolutional layer of the CNN of the CNN-based filter, and Alpha is a fixed parameter.

[0236]

[0222] Aspect 18C: A device described in aspect 17C, wherein the CNN-based filter includes multiple convolutional layers and the value of Alpha is the same for each of the convolutional layers.

[0237]

[0223] Aspect 19C: A device described in aspect 17C, wherein the CNN-based filter includes multiple convolutional layers and the value of Alpha is different for two or more of the convolutional layers.

[0238]

[0224] Aspect 20C: A device described in aspects 17C to 19C, wherein the one or more processors are further configured to determine a value for Alpha based on one or more syntax elements signaled in a bitstream comprising an encoded representation of the video data.

[0239]

[0225] Aspect 21C: The device of aspect 20C, wherein the value of Alpha is signaled in the bitstream.

[0240]

[0226] Aspect 22C: A device described in aspect 20C, wherein an index of Alpha within a predefined set is signaled in the bitstream, and wherein determining the value of Alpha comprises determining the value of Alpha in the predefined set to which the index of Alpha corresponds.

[0241]

[0227] Aspect 23C: A device described in any of aspects 17C to 18C or 20C to 22C, wherein the CNN-based filter includes multiple convolutional layers, and wherein one or more processors are further configured to determine a single value of Alpha for use in a LeakyReLU activation function applied to output values ​​of the convolutional layers, and to encode, in a bitstream comprising an encoded representation of the video data, one or more syntax elements that provide information required by a video decoder to derive the value of Alpha.

[0242]

[0228] Aspect 24C: A device described in any of aspects 17C or 18C to 22C, wherein the CNN includes multiple convolutional layers, and wherein one or more processors are further configured to determine multiple values ​​of Alpha for use in a LeakyReLU activation function applied to output values ​​of two or more different convolutional layers of the multiple convolutional layers, and encode, in a bitstream comprising an encoded representation of the video data, one or more syntax elements that provide information required by a video decoder to derive the values ​​of Alpha.

[0243]

[0229] Aspect 25C: The device of any of aspects 16C to 24C, further comprising a display configured to display the decoded video data.

[0244] Aspect 26C: The device of any of aspects 16C to 25C, wherein the device comprises one or more of a camera, a computer, a mobile device, a broadcast receiver device, or a set-top box.

[0245]

[0231] Aspect 27C: A device according to any one of aspects 16C to 26C, wherein the device comprises a video decoder.

[0246]

[0232] Aspect 28C: A device described in any of aspects 16C to 27C, wherein the device comprises a video encoder.

[0247]

[0233] Aspect 29C: A device for coding video data includes means for reconstructing blocks of video data and means for applying a convolutional neural network (CNN)-based filter to the reconstructed blocks, wherein the CNN-based filter uses a LeakyReLU activation function.

[0248] Aspect 29C': A device according to aspect 29C, comprising means for carrying out the method according to any one of claims 1C to 11C.

[0249]

[0235] Aspect 30C: A computer-readable storage medium having stored thereon instructions that, when executed, cause one or more processors to reconstruct blocks of video data and apply a convolutional neural network (CNN)-based filter to the reconstructed blocks, wherein the CNN-based filter uses a LeakyReLU activation function.

[0250] Aspect 30C': A computer-readable storage medium according to aspect 26C, comprising means for performing the method according to any one of claims 1C to 15C.

[0251]

[0237] In accordance with the above examples, it should be recognized that some acts or events of any of the techniques described herein may be performed in a different sequence, added, merged, or entirely excluded (e.g., not all described acts or events may be required to practice the techniques). Moreover, in some examples, acts or events may be performed simultaneously rather than sequentially, for example, through multithreaded processing, interrupt processing, or multiple processors.

[0252] In one or more examples, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions may be stored on or transmitted via a computer-readable medium as one or more instructions or code and executed by a hardware-based processing unit. Computer-readable media may include computer-readable storage media, which correspond to tangible media such as data storage media, or may include communication media, including any medium that enables transfer of a computer program from one place to another, for example, according to a communications protocol. In this manner, computer-readable media may generally correspond to (1) tangible computer-readable storage media that are non-transitory, or (2) communication media such as a signal or carrier wave. Data storage media may be any available medium that can be accessed by one or more computers or one or more processors to retrieve instructions, code, and / or data structures for implementing the techniques described in this disclosure. A computer program product may include a computer-readable medium.

[0253]

[0239] By way of example, and not limitation, such computer-readable storage media may comprise RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic 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. Also, any connection is properly termed 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 technologies such as infrared, radio, and microwave, the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are 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 transitory media, but instead cover non-transitory, tangible storage media. As used herein, disk and disc include compact discs (CDs), laser discs, optical discs, digital versatile discs (DVDs), floppy discs, and Blu-ray discs, where disks typically reproduce data magnetically and discs reproduce data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0254]

[0240] The instructions may be executed by one or more processors, such as one or more DSPs, general-purpose microprocessors, ASICs, FPGAs, or other equivalent integrated circuits or discrete logic circuits. Accordingly, the terms "processor" and "processing circuitry" as used herein may refer to any of the above structures or any other structure suitable for implementing the techniques described herein. Furthermore, in some aspects, the functionality described herein may be provided in dedicated hardware and / or software modules configured for encoding and decoding, or incorporated into a combined codec. Also, the techniques may be fully implemented in one or more circuits or logic elements.

[0255] The techniques of this disclosure may be implemented in a wide variety of devices or apparatuses, including wireless handsets, integrated circuits (ICs), or sets of ICs (e.g., chipsets). Although various components, modules, or units have been described in this disclosure to highlight functional aspects of devices configured to implement the disclosed techniques, those components, modules, or units do not necessarily require realization by different hardware units. Rather, as described above, the various units may be combined in a codec hardware unit or provided by a collection of interoperable hardware units, including one or more processors described above, along with suitable software and / or firmware.

[0256]

[0242] Various examples have been described. These and other examples are within the scope of the following claims. The inventions described in the claims of the present application as originally filed are set forth below. [C1] 1. A method for encoding or decoding video data, the method comprising: reconstructing the blocks of video data; applying a convolutional neural network (CNN)-based filter to the reconstructed block, wherein the CNN-based filter uses a LeakyReLU activation function; A method comprising: [C2] The LeakyReLU activation function is

number

number

Claims

1. 1. A method for encoding video data, the method comprising: reconstructing the blocks of video data; determining a plurality of values ​​for an Alpha parameter; applying a convolutional neural network (CNN) based filter to the reconstructed block; and the CNN-based filter uses a leaky rectified linear unit (LeakyReLU) activation function; The LeakyReLU activation function is [Equation 1] where y is the output value of the convolutional layer of the CNN of the CNN-based filter, and the CNN includes multiple convolutional layers; applying the CNN-based filter to the reconstructed block comprises using different values ​​of the Alpha parameter from the plurality of values ​​of the Alpha parameter in two or more different convolutional layers of the plurality of convolutional layers. encoding, in a bitstream comprising the coded representation of the video data, one or more syntax elements that provide information for deriving the values ​​of the Alpha parameter at a video decoder; A method comprising:

2. The method of claim 1 , wherein the one or more syntax elements directly specify the value of the Alpha parameter.

3. The method of claim 1 , wherein the one or more syntax elements specify an index that corresponds to the value of the Alpha parameter within a predefined set of indexes.

4. 1. A device for encoding video data, said device comprising: a memory for storing the video data; and one or more processors implemented in circuitry, the one or more processors comprising: reconstructing the blocks of video data; determining a plurality of values ​​for an Alpha parameter; applying a convolutional neural network (CNN) based filter to the reconstructed block; and the CNN-based filter uses a leaky rectified linear unit (LeakyReLU) activation function; The LeakyReLU activation function is [Equation 2] where y is the output value of the convolutional layer of the CNN of the CNN-based filter, and the CNN includes multiple convolutional layers. wherein the one or more processors are configured to: using different values ​​of the Alpha parameter from the plurality of values ​​of the Alpha parameter in two or more different convolutional layers of the plurality of convolutional layers as part of applying the CNN-based filter to the reconstructed block; encoding, in a bitstream comprising the coded representation of the video data, one or more syntax elements that provide information for deriving the values ​​of the Alpha parameter at a video decoder; A device configured to:

5. The device of claim 4 , wherein the device comprises one or more of a camera, a computer, a mobile device, a broadcast receiver device, or a set-top box.

6. The device of claim 4 , wherein the device comprises a video encoder.

7. A computer readable storage medium having stored thereon instructions that, when executed, cause one or more processors to perform the method of any one of claims 1 to 3.

Images