1 Scope
2 Normative references
     2.1 Identical Recommendations | International Standards
     2.2 Paired Recommendations | International Standards equivalent in technical content
     2.3 Additional references
3 Definitions
4 Abbreviations
5 Conventions
     5.1 General
     5.2 Arithmetic operators
     5.3 Logical operators
     5.4 Relational operators
     5.5 Bit-wise operators
     5.6 Assignment operators
     5.7 Range notation
     5.8 Mathematical functions
     5.9 Order of operation precedence
     5.10 Variables, syntax elements and tables
     5.11 Text description of logical operations
     5.12 Processes
6 Image representation, tensor operations and neural network
     6.1 Image representation
     6.2 Tensor operations and neural network design elements
          6.2.1 Average pooling 8
          6.2.2 Batch normalization
          6.2.3 Bicubic upsampling
          6.2.4 Concatenation of two tensors
          6.2.5 Concatenation of three tensors
          6.2.6 Convolution layer
          6.2.7 Convolution-based attention block
          6.2.8 Cropping layer
          6.2.9 Deinteger function
          6.2.10 Downsampling layer
          6.2.11 Downshuffle operation
          6.2.12 Grouped convolution layer
          6.2.13 Layer normalization
          6.2.14 Latent combine block
          6.2.15 Lightweight residual block
          6.2.16 Matrix transpose
          6.2.17 Matrix multiplication
          6.2.18 Multiplication of tensors by a constant
          6.2.19 Nearest neighbour downsampling
          6.2.20 Padding layer
          6.2.21 Per element absolute value
          6.2.22 Per element addition of tensors
          6.2.23 Per element multiplication of tensors
          6.2.24 Quantized convolution
          6.2.25 Quantized transposed convolution
          6.2.26 Rectified exponential unit
          6.2.27 Rectified linear unit
          6.2.28 Rectified linear unit 6
          6.2.29 Reshape
          6.2.30 Residual activation unit
          6.2.31 Residual block
          6.2.32 Shuffle layer
          6.2.33 Sigmoid
          6.2.34 Soft-max
          6.2.35 Tensor chunk
          6.2.36 Tensor normalization
          6.2.37 Transformer-based attention block
          6.2.38 Transformer-based attention module
          6.2.39 Transposed convolution
          6.2.40 Unshuffle layer
          6.2.41 Upsampling layer
          6.2.42 Upshuffle operation
          6.2.43 Zero padding
7 Overview
     7.1 Learnable neural network models
     7.2 An overview on the encoding and decoding processes
     7.3 Functional overview on the encoding processes
     7.4 Functional overview on the decoding process
     7.5 Encoder requirements
     7.6 Decoder requirements
     7.7 Models and weights
8 Syntax and semantics
     8.1 Method of specifying syntax in tabular form
     8.2 Specification of the parsing process
          8.2.1 Specification of syntax functions and descriptors
          8.2.2 Parsing process for k-th order Exp-Golomb codes
          8.2.3 Mapping process for signed Exp-Golomb codes
9 Colour component separation
     9.1 General
     9.2 Conversion to coding format
     9.3 Example of conversion to coding colour space for RGB input
     9.4 Example of conversion to coding colour space for YCbCr input
     9.5 Downsampling for secondary component
     9.6 Upsampling of secondary component
     9.7 Conversion to output picture format
     9.8 Conversion from coding colour space to output image colour space
10 Latent domain tiles
     10.1 Neural network modules processing with tiles
          10.1.1 Tiling processing
          10.1.2 Tiles size and location computation
          10.1.3 Tiling of input tensor
          10.1.4 Merging tiles into output tensor
          10.1.5 Overlap amount determination for tile
11 Synthesis transform
     11.1 Analysis transform network (informative)
     11.2 Synthesis transform with latent domain tiles
     11.3 Synthesis transform network
12 Codestream structure and entropy coder
     12.1 Codestream layout
     12.2 Substream syntax
          12.2.1 Substream syntax structure
          12.2.2 Semantics of substreams
     12.3 Picture header
          12.3.1 Picture header syntax
          12.3.2 Picture header semantics
     12.4 Residual and hyper tensor coding
          12.4.1 Hyper tensor decoding
               12.4.1.1 Hyper tensor decoding process
               12.4.1.2 Syntax table for hyper tensor
               12.4.1.3 Semantics hyper tensor stream
               12.4.1.4 Decoding of hyper tensor substream
               12.4.1.5 Three-dimensional hyper-tensor reconstruction
          12.4.2 Quality map information decoder
               12.4.2.1 Quality map information decoder process
               12.4.2.2 Syntax table for quality map information tensor decoding
               12.4.2.3 Semantics of quality map information tensor
               12.4.2.4 Three-dimensional quality map tensor reconstruction
          12.4.3 Residual data decoding
               12.4.3.1 Residual data decoding process
               12.4.3.2 Syntax table for residual tensor decoding
               12.4.3.3 Semantics of residual tensor
     12.5 Entropy coder me-tANS
          12.5.1 General
          12.5.2 Initialization of the codestream segment
          12.5.3 Data decoding
          12.5.4 Pre-calculation steps for me-tANS
               12.5.4.1 Pre-calculation for cumulative distribution function tables
               12.5.4.2 Pre-calculation for decoding transition tables
     12.6 Tools header
          12.6.1 General
          12.6.2 Tools header syntax table
          12.6.3 Tools information semantics
     12.7 User defined information
          12.7.1 General
          12.7.2 User defined information syntax table
          12.7.3 User defined information semantics
     12.8 Rendering information
          12.8.1 General
          12.8.2 Rendering information syntax
               12.8.2.1 CICP rendering information
               12.8.2.2 Mastering display colour volume
               12.8.2.3 Content light level information
               12.8.2.4 Dynamic metadata
          12.8.3 Rendering information semantics
13 Entropy parameters decoding
     13.1 General
     13.2 Hyper encoder
          13.2.1 Hyper encoder process
          13.2.2 Single component hyper encoder
     13.3 Hyper scale decoder
          13.3.1 Hyper scale decoder process
          13.3.2 Single-component hyper scale decoder
     13.4 Sigma scale
     13.5 Adaptive sigma scale
     13.6 Sigma quantization
14 Latent domain prediction and residual
     14.1 General
     14.2 Hyper decoder
          14.2.1 Hyper decoder process
          14.2.2 Hyper decoder with tiling
          14.2.3 Single-component hyper decoder
     14.3 Latent tensor reconstruction
          14.3.1 Latent tensor reconstruction with tiling
          14.3.2 Latent reconstruction
          14.3.3 Multistage context modelling
               14.3.3.1 Stage 0 of multistage context modelling (MCM(0))
               14.3.3.2 Stage 1 of multistage context modelling (MCM(1))
               14.3.3.3 Stage 2 of multistage context modelling (MCM(2))
               14.3.3.4 Stage 3 of multistage context modelling (MCM(3))
15 Variable rate support
     15.1 General
     15.2 Control parameters and gain tensor derivation
     15.3 Gain unit
     15.4 Inverse gain unit
16 Mask and scale tools
     16.1 General
     16.2 Residual and variance scaling (RVS)
          16.2.1 RVS process
          16.2.2 RVS scaling tensor generation
          16.2.3 Precalculation of RVS tables
          16.2.4 Residual scale
          16.2.5 Inverse residual scale
     16.3 SKIP mode
          16.3.1 SKIP mode process
          16.3.2 Cube flag generation
          16.3.3 SKIP mask generation
     16.4 Latent scale before synthesis (LSBS)
          16.4.1 LSBS process
          16.4.2 Pre-calculation of LSBS tables
          16.4.3 Decoder LSBS operation
17 Enhancement filter technologies
     17.1 General
     17.2 Secondary component re-sampling in absence of adaptive linear filter
     17.3 Adaptive linear filter
          17.3.1 Adaptive linear filter process
          17.3.2 Adaptive linear filter parameters updating process
          17.3.3 Adaptive linear filter tiling process
     17.4 Inter channel correlation information filter
          17.4.1 Inter channel correlation information filter process
          17.4.2 Inter Channel Correlation Information filter for one tile
     17.5 Non-linear chroma enhancement filter
          17.5.1 Non-linear chroma enhancement filter process
          17.5.2 Non-linear filter parameters tiling process
     17.6 Luma edge filtering (LEF) filter
18 Constant tables
     18.1 General
          18.1.1 Probability distribution tables for hyper-tensor coding
               18.1.1.1 Mapping tables for hyper stream decoding
               18.1.1.2 Cumulative distribution function tables for hyper stream decoding
               18.1.1.3 Probability distribution function tables for hyper stream decoding
          18.1.2 Probability distribution tables for residual coding
     18.2 Trained model parameters
     18.3 Entropy coder constants
Bibliography