WO2025212980A1 - Method, apparatus, and medium for visual data processing - Google Patents

Method, apparatus, and medium for visual data processing

Info

Publication number
WO2025212980A1
WO2025212980A1 PCT/US2025/023092 US2025023092W WO2025212980A1 WO 2025212980 A1 WO2025212980 A1 WO 2025212980A1 US 2025023092 W US2025023092 W US 2025023092W WO 2025212980 A1 WO2025212980 A1 WO 2025212980A1
Authority
WO
WIPO (PCT)
Prior art keywords
parameter
visual data
codestream
model
indication
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
PCT/US2025/023092
Other languages
French (fr)
Inventor
Semih Esenlik
Zhaobin Zhang
Ye-Kui Wang
Kai Zhang
Li Zhang
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
ByteDance Inc
Original Assignee
ByteDance Inc
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by ByteDance Inc filed Critical ByteDance Inc
Publication of WO2025212980A1 publication Critical patent/WO2025212980A1/en
Pending legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G06N3/0455—Auto-encoder networks; Encoder-decoder networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/047—Probabilistic or stochastic networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
    • G06N3/063—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using electronic means
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/088—Non-supervised learning, e.g. competitive learning
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/115—Selection of the code volume for a coding unit prior to coding
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
    • H04N19/17—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object
    • H04N19/172—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object the region being a picture, frame or field
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/70—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by syntax aspects related to video coding, e.g. related to compression standards

Definitions

  • Embodiments of the present disclosure relates generally to visual data processing techniques, and more particularly, to neural network-based visual data coding.
  • BACKGROUND [0002] The past decade has witnessed the rapid development of deep learning in a variety of areas, especially in computer vision and image processing. Neural network was invented originally with the interdisciplinary research of neuroscience and mathematics. It has shown strong capabilities in the context of non-linear transform and classification. Neural network- based image/video compression technology has gained significant progress during the past half decade.
  • At least one indication associated with a usage of a variable rate coding process in the conversion is signaled in the bitstream.
  • the proposed method can better support the application of variable rate coding process. Thereby, the coding flexibility can be improved.
  • an apparatus for visual data processing is proposed.
  • the apparatus comprises a processor and a non-transitory memory with instructions thereon. The instructions upon execution by the processor, cause the processor to perform a method in accordance with the first aspect of the present disclosure.
  • a non-transitory computer-readable storage medium is proposed.
  • the non-transitory computer-readable storage medium stores instructions that cause a 1 F1250625PCT processor to perform a method in accordance with the first aspect of the present disclosure.
  • a fourth aspect another non-transitory computer-readable recording medium is proposed.
  • the non-transitory computer-readable recording medium stores a codestream of visual data which is generated by a method performed by an apparatus for visual data processing.
  • the method comprises: performing a conversion from the visual data to the codestream with a neural network (NN)-based model, wherein the codestream comprises at least one indication associated with a usage of a variable rate coding process in the conversion.
  • NN neural network
  • a method for storing a codestream of visual data is proposed.
  • the method comprises: performing a conversion from the visual data to the codestream with a neural network (NN)-based model; and storing the codestream in a non-transitory computer- readable recording medium, wherein the codestream comprises at least one indication associated with a usage of a variable rate coding process in the conversion.
  • NN neural network
  • Fig. 1A illustrates a block diagram that illustrates an example visual data coding system in accordance with some embodiments of the present disclosure
  • Fig. 1B is a schematic diagram illustrating an example transform coding scheme
  • Fig. 2 illustrates example latent representations of an image
  • Fig. 1A illustrates a block diagram that illustrates an example visual data coding system in accordance with some embodiments of the present disclosure
  • Fig. 1B is a schematic diagram illustrating an example transform coding scheme
  • Fig. 2 illustrates example latent representations of an image
  • Fig. 1A illustrates a block diagram that illustrates an example visual data coding system in accordance with some embodiments of the present disclosure
  • Fig. 1B is a schematic diagram illustrating an example transform coding scheme
  • Fig. 2 illustrates example latent representations of an image
  • FIG. 3 is a schematic diagram illustrating an example autoencoder implementing a hyperprior model
  • Fig. 4 is a schematic diagram illustrating an example combined model configured to jointly optimize a context model along with a hyperprior and the autoencoder
  • Fig. 5 illustrates an example encoding process
  • Fig. 6 illustrates an example decoding process
  • Fig. 7 illustrates an example decoding process according to some embodiments of the present disclosure
  • Fig. 8 illustrates an example learning-based image codec architecture
  • 2 F1250625PCT [0021]
  • Fig. 9 illustrates an example synthesis transform for learning based image coding
  • FIG. 10 illustrates an example leaky Rectified Linear Unit (ReLU) activation function
  • FIG. 11 illustrates an example ReLU activation function
  • Fig. 12 illustrates an example down-shuffle operation
  • Fig. 13 illustrates an example up-shuffle operation
  • Fig. 14 illustrates latent tiles in synthesis transform
  • Fig. 15 illustrates a bitstream layout
  • Fig. 16 illustrates an example decoder structure
  • Fig. 17 illustrates an example hyper scale decoder
  • Fig. 18 illustrates an example hyper decoder
  • FIG. 19 illustrates a diagram of an example multistage context modelling (MCM) structure
  • MCM multistage context modelling
  • Fig. 20 illustrates an example implementation of primary component guided adaptive up-sampling filter
  • Fig. 21 illustrates a luma edge filtering (LEF) general process
  • Fig. 22 illustrates an MCM model to obtain cube_flag
  • Fig. 23 illustrates a bitstream structure in accordance with embodiments of the present disclosure
  • Fig. 24 illustrates another bitstream structure in accordance with embodiments of the present disclosure
  • FIG. 25 illustrates a further bitstream structure in accordance with embodiments of the present disclosure
  • Fig.26 illustrates a still further bitstream structure in accordance with embodiments of the present disclosure
  • Fig. 27 illustrates an example of reshaping a one-dimensional (1D) array of cube_flag into a two dimensional (2D) array
  • Fig. 28 illustrates a flowchart of a method for visual data processing in accordance with embodiments of the present disclosure
  • Fig. 29 illustrates a block diagram of a computing device in which various embodiments of the present disclosure can be implemented.
  • the same or similar reference numerals usually refer to the same or similar elements.
  • references in the present disclosure to “one embodiment,” “an embodiment,” “an example embodiment,” and the like indicate that the embodiment described may include a particular feature, structure, or characteristic, but it is not necessary that every embodiment includes the particular feature, structure, or characteristic. Moreover, such phrases are not necessarily referring to the same embodiment. Further, when a particular feature, structure, or characteristic is described in connection with an example embodiment, it is submitted that it is within the knowledge of one skilled in the art to affect such feature, structure, or characteristic in connection with other embodiments whether or not explicitly described. [0046] It shall be understood that although the terms “first” and “second” etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another.
  • first element could be termed a second element, and similarly, a second element could be termed a first element, without departing from the scope of example embodiments.
  • the term “and/or” includes any and all combinations of one or more of the listed terms. [0047]
  • the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments. As used herein, the singular forms “a”, “an” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise.
  • FIG. 1A is a block diagram that illustrates an example visual data coding system 100 that may utilize the techniques of this disclosure.
  • the visual data coding system 100 may include a source device 110 and a destination device 120.
  • the source device 110 can be also referred to as a visual data encoding device
  • the destination device 120 4 F1250625PCT can be also referred to as a visual data decoding device.
  • the source device 110 can be configured to generate encoded visual data and the destination device 120 can be configured to decode the encoded visual data generated by the source device 110.
  • the source device 110 may include a visual data source 112, a visual data encoder 114, and an input/output (I/O) interface 116.
  • the visual data source 112 may include a source such as a visual data capture device. Examples of the visual data capture device include, but are not limited to, an interface to receive visual data from a visual data provider, a computer graphics system for generating visual data, and/or a combination thereof.
  • the visual data may comprise one or more pictures of a video or one or more images.
  • the visual data encoder 114 encodes the visual data from the visual data source 112 to generate a bitstream.
  • the bitstream may include a sequence of bits that form a coded representation of the visual data.
  • the bitstream may include coded pictures and associated visual data.
  • the coded picture is a coded representation of a picture.
  • the associated visual data may include sequence parameter sets, picture parameter sets, and other syntax structures.
  • the I/O interface 116 may include a modulator/demodulator and/or a transmitter.
  • the encoded visual data may be transmitted directly to destination device 120 via the I/O interface 116 through the network 130A.
  • the encoded visual data may also be stored onto a storage medium/server 130B for access by destination device 120.
  • the destination device 120 may include an I/O interface 126, a visual data decoder 124, and a display device 122.
  • the I/O interface 126 may include a receiver and/or a modem.
  • the I/O interface 126 may acquire encoded visual data from the source device 110 or the storage medium/server 130B.
  • the visual data decoder 124 may decode the encoded visual data.
  • the display device 122 may display the decoded visual data to a user.
  • the display device 122 may be integrated with the destination device 120, or may be external to the destination device 120 which is configured to interface with an external display device.
  • the visual data encoder 114 and the visual data decoder 124 may operate according to a visual data coding standard, such as video coding standard or still picture coding standard and other current and/or further standards.
  • a visual data coding standard such as video coding standard or still picture coding standard and other current and/or further standards.
  • the term visual data processing encompasses visual data coding or compression, visual data decoding or decompression and visual data transcoding in which visual data are represented from one compressed format into another compressed format or at a different compressed bitrate.
  • NN neural network
  • the present disclosure is related to neural network (NN)-based image and video coding. Specifically, it is related to the method of signaling tool header of skip mode in support of regional accessibility, wherein regional accessibility refers to the capability of correctly decoding only a regional part of an image (also referred to as a picture) or a video.
  • this disclosure is related to a neural network-based image and video compression method comprising modification of components of an image using convolution layers.
  • the weights of the convolution layer is included in the bitstream.
  • Image/video compression (also referred to as image/video coding) usually refers to the computing technology that compresses image/video into binary code to facilitate storage and transmission.
  • the binary codes may or may not support losslessly reconstructing the original image/video, termed lossless compression and lossy compression.
  • Neural network- based video compression is in two flavors, neural network-based coding tools and end-to-end neural network-based video compression.
  • the former is embedded into existing classical video codecs as coding tools and only serves as part of the framework, while the latter is a separate framework developed based on neural networks without depending on classical video codecs.
  • a series of classical video coding standards have been developed to accommodate the increasing visual content.
  • the international standardization organizations ISO/IEC has two expert groups namely Joint Photographic Experts Group (JPEG) and Moving Picture Experts Group (MPEG), and ITU-T also has its own Video Coding Experts Group (VCEG) which is for standardization of image/video coding technology.
  • JPEG Joint Photographic Experts Group
  • MPEG Moving Picture Experts Group
  • VCEG Video Coding Experts Group
  • Neural networks also known as artificial neural networks (ANN) are the computational models used in machine learning technology which are usually composed of multiple processing layers and each layer is composed of multiple simple but non-linear basic computational units.
  • ANN artificial neural networks
  • One benefit of such deep networks is believed to be the capacity for processing data with multiple 7 F1250625PCT levels of abstraction and converting data into different kinds of representations.
  • neural networks for image compression methods can be classified in two categories, i.e., pixel probability modeling and auto-encoder. The former one belongs to the predictive coding strategy, while the latter one is the transform-based solution. Sometimes, these two methods are combined together in literature.
  • one way to model ⁇ is to predict pixel probabilities one by one in a raster scan order based on previous observations, where ⁇ is an image.
  • ⁇ is an image.
  • ⁇ , ... , ⁇ (1) where is also known as the context of the current pixel.
  • ⁇ is a pre-defined constant controlling the range of the context. It should be noted that the condition may also take the sample values of other color components into consideration.
  • the current G sample may be coded according to previously coded pixels and the current R sample, while for coding the current B sample, the previously coded pixels and the current R and G samples may also be taken into consideration.
  • Neural networks were originally introduced for computer vision tasks and have been proven to be effective in regression and classification problems. Therefore, it has been proposed using neural 8 F1250625PCT networks to estimate the probability of ⁇ given its context ⁇ , ⁇ , ... , ⁇ . Most of the methods directly model the probability distribution in the pixel domain. Some researchers also attempt to model the probability distribution as a conditional one upon explicit or latent representations.
  • the latent samples are modeled as gaussian distribution or gaussian mixture models (not limited to).
  • the context model and hyper prior are jointly used to estimate the probability distribution of the latent samples. Since a gaussian distribution can be defined by a mean and a variance (aka sigma or scale), the joint model is used to estimate the mean and variance (denoted as ⁇ and ⁇ ).
  • G-VAE Gained variational autoencoders
  • neural network-based image/video compression methodologies need to train multiple models to adapt to different rates.
  • Gained variational autoencoders is the variational autoencoder with a pair of gain units , which is designed to achieve continuously variable rate adaptation using a single model. It comprises of a pair of gain units, which are typically inserted to the output of encoder and input of decoder.
  • the output of the encoder is defined as the latent r epresentation ⁇ ⁇ ⁇ , where ⁇ , h, ⁇ represent the number of channels, the height and width of the latent
  • Each channel of the latent representation is denoted as ⁇ ⁇ ⁇ , where ⁇ ⁇ 0, 1, ... , ⁇ ⁇ 1.
  • a pair of gain units include a gain matrix ⁇ ⁇ ⁇ and inverse gain matrix, where ⁇ is the number of gain vectors.
  • the gain vector can be denoted as ⁇ ⁇ ⁇ , ⁇ , ... , ⁇ , ⁇ ⁇ ⁇ where ⁇ denotes the index of the gain vectors in the gain
  • ⁇ denotes the index of the gain vectors in the gain
  • the motivation of gain matrix is similar to the quantization table in JPEG by controlling the quantization loss based on the characteristics of different channels.
  • each channel is multiplied with the corresponding value in a gain vector. 12 F1250625PCT ⁇ ⁇ ⁇ ⁇ ⁇ where ⁇ is channel-wise multiplication, i.e., ⁇ ⁇ ⁇ ⁇ ⁇ , and ⁇ is the ⁇ -th gain value in the gain vector ⁇ .
  • the in ⁇ ⁇ verse gain matrix used at the decoder side can be denoted as ⁇ ′ ⁇ ⁇ , which consists of ⁇ inverse gain vectors, i.e., ⁇ ′ ⁇ ⁇ , ⁇ , ... , ⁇ , ⁇ ⁇ ⁇ .
  • the inverse gain process is expressed as: ′ ⁇ ′ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ′ where ⁇ is the decoded quantized latent representation and ⁇ ⁇ is the inversely gained quantized latent representation, which will be fed into the synthesis network.
  • interpolation is used between vectors.
  • the interpolated gain vector can be obtained via the following equations. ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ′ ′ ′ where ⁇ ⁇ ⁇ is an interpolation the corresponding bit rate of the generated gain vector pair. Since ⁇ is a real number, an arbitrary bit rate between the given two gain vector pairs can be achieved. 2.3.6
  • the encoding process using joint auto-regressive hyper prior model The fig 4. corresponds to the state of the art compression method. In this section and the next, the encoding and decoding processes will be described separately.
  • the Fig. 5 depicts the encoding process.
  • the latent ⁇ is input to hyper encoder, which outputs the hyper latent (denoted by ⁇ ).
  • the hyper latent is then quantized ( ⁇ ) and a second bitstream (bits2) is generated using arithmetic encoding (AE) module.
  • AE arithmetic encoding
  • the factorized entropy module generates the probability distribution, that is used to encode the quantized hyper latent into bitstream.
  • the quantized hyper latent includes information about the probability distribution of the quantized latent ( ⁇ ⁇ ⁇ ). 13 F1250625PCT
  • the Entropy Parameters subnetwork generates the probability distribution estimations, that are used to encode the quantized latent ⁇ .
  • the information that is generated by the Entropy Parameters typically include a mean ⁇ and scale (or variance) ⁇ parameters, that are together used to obtain a gaussian probability distribution.
  • a gaussian distribution of a random variable x is defined as ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ wherein the parameter ⁇ is the mean or expectation of the distribution (and also its median and mode), while the parameter ⁇ is its standard deviation (or variance, or scale).
  • the mean and the variance need to be determined.
  • the entropy parameters module are used to estimate the mean and the variance values.
  • Fig. 5 The analysis transform that converts the input image into latent representation is also called an encoder (or auto-encoder). 2.3.7 The decoding process using joint auto-regressive hyper prior model
  • the Fig.6 depicts the decoding process separately.
  • the decoder first receives the first bitstream (bits1) and the second bitstream (bits2) that are generated by a corresponding encoder.
  • the bits2 is first decoded by the arithmetic decoding (AD) module by utilizing the probability distributions generated by the 14 F1250625PCT factorized entropy subnetwork.
  • AD arithmetic decoding
  • autoregressive model (the context model) is inherently serial, and therefore cannot be sped up using techniques such as parallelization.
  • the fully reconstructed quantized latent ⁇ ⁇ ⁇ is input to the synthesis transform (denoted as decoder in Fig.6) module to obtain the reconstructed image.
  • decoder the all of the elements in Fig. 6 are collectively called decoder.
  • the synthesis transform that converts the quantized latent into reconstructed image is also called a decoder (or auto-decoder).
  • the lossless methods can achieve compression ratio of about 1.5 to 3 for natural images, which is clearly below requirement. Therefore, lossy compression is developed to achieve further compression ratio, but at the cost of incurred distortion.
  • the distortion can be measured by calculating the average squared difference between the original image and the reconstructed image, i.e., mean-squared-error (MSE).
  • MSE mean-squared-error
  • ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ (4) Accordingly, the quality of the 16 F1250625PCT measured by peak signal-to-noise ratio (PSNR): ⁇ ⁇ ⁇ 10 ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ (5) where ⁇ is the quality evaluation metrics SSIM).
  • the luma and chroma components of an image can be decoded using separate subnetworks.
  • the luma component of the image is processed by the subnetwoks “Synthesis”, “Prediction fusion”, “Mask Conv”, “Hyper Decoder”, “Hyper scale decoder” etc.
  • the chroma components are processed by the subnetworks: “Synthesis UV”, “Prediction fusion UV”, “Mask Conv UV”, “Hyper Decoder UV”, “Hyper scale decoder UV” etc.
  • a benefit of the above separate processing is that the computational complexity of the processing of an image is reduced by application of separate processing.
  • the computational complexity is proportional to the square of the number of feature maps. If the number of total feature maps is equal to 192 for example, computational complexity will be proportional to 192x192. On the other hand if the feature maps are divided into 128 for luma and 64 for chroma (in the case of separate processing), the computational complexity is proportional to 128x128 + 64x64, which corresponds to a reduction in complexity by 45%.
  • the separate processing of luma and chroma components of an image does not result in a prohibitive reduction in performance, as the correlation between the luma and chroma components are typically very small.
  • the processing (Decoding process) in Fig.7 can be explained below: 1.
  • the factorized entropy model is used to decode the quantized latents for luma and chroma, i.e., ⁇ and ⁇ ⁇ in Fig.7. 2.
  • the probability parameters (e.g. variance) generated by the second network are used to generate a quantized residual latent by performing the arithmetic decoding process. 17 F1250625PCT 3.
  • the quantized residual latent is inversely gained with the inverse gain unit (iGain) as shown i n orange color in Fig. 7.
  • the outputs of the inverse gain units are denoted as ⁇ and ⁇ for luma and chroma components, respectively. 4.
  • a For the luma component, the following steps are performed in a loop until all elements of ⁇ ⁇ are obtained: a. A first subnetwork is used to estimate a mean value parameter of a quantized latent ( ⁇ ), using the already obtained samples of ⁇ ⁇ . b . The quantized residual latent ⁇ and the mean value are used to obtain the next element of ⁇ ⁇ . 5. After all of the samples of ⁇ ⁇ are obtained, a synthesis transform can be applied to obtain the reconstructed image. 6. For chroma component, step 4 and 5 are the same but with a separate set of networks. 7. The decoded luma component is used as additional information to obtain the chroma component.
  • the Inter Channel Correlation Information filter sub-network (ICCI) is used for chroma component restoration.
  • the luma is fed into the ICCI sub- network as additional information to assist the chroma component decoding.
  • Adaptive color transform (ACT) is performed after the luma and chroma components are reconstructed.
  • the module named ICCI is a neural-network based postprocessing module.
  • the example embodiments of the present disclosure are not limited to the UCCI subnetwork, any other neural network based postprocessing module might also be used.
  • An exemplary implementation of some example embodiments of the present disclosure is depicted in Fig. 7 (the decoding process).
  • the framework comprises two branches for luma and chroma components respectively.
  • the first subnetwork comprises the context, prediction and optionally the hyper decoder modules.
  • the second network comprises the hyper scale decoder module.
  • the quantized hyper latent are ⁇ and ⁇ ⁇ .
  • the arithmetic decoding process generates the quantized residual latents, which are further fed into the iGain units to obtain the gained quantized residual latents ⁇ and ⁇ .
  • a recursive prediction operation is performed to obtain the latent ⁇ ⁇ and ⁇ ⁇ .
  • the following steps describe how to obtain the samples of latent ⁇ ⁇ : , ⁇ , ⁇ , and the chroma component is processed in the same way but with different networks. 1.
  • An autoregressive context module is used to generate first input of a prediction module u sing the samples ⁇ ⁇ : , ⁇ , ⁇ where the (m, n) pair are the indices of the samples of the latent that are already obtained.
  • the second input of the prediction module is obtained by using a hyper decoder 18 F1250625PCT and a quantized hyper latent ⁇ ⁇ ⁇ .
  • the prediction module uses the first input and the second input, the prediction module generates the mean value ⁇ ⁇ : , ⁇ , ⁇ . 4.
  • Whether to and/or how to apply at least one method disclosed in the document may be signaled from the encoder to the decoder, e.g. in the bitstream. Alternatively, whether to and/or how to apply at least one method disclosed in the document may be determined by the decoder based on coding information, such as dimensions, color format, etc.
  • the modules named MS1, MS2 or MS3+O might be included in the processing flow. The said modules might perform an operation to their input by multiplying the input with a scalar or adding an adding an additive component to the input to obtain the output. The scalar or the additive component that are used by the said modules might be indicated in a bitstream.
  • the module named RD or the module named AD in Fig.7 might be an entropy decoding module. It might be a range decoder or an arithmetic decoder or the like.
  • the example embodiments of the present disclosure described herein are not limited to the specific combination of the units exemplified in Fig. 7. Some of the modules might be missing and some of the modules might be displaced in processing order. Also additional modules might be included. For example: 1.
  • the ICCI module might be removed. In that case the output of the Synthesis module and the Synthesis UV module might be combined by means of another module, that might be based on neural networks. 2.
  • One or more of the modules named MS1, MS2 or MS3+O might be removed.
  • the core of the proposed solution is not affected by the removing of one or more of the said scaling and adding modules.
  • FIG. 7 other operations that are performed during the processing of the luma and chroma components are also indicated using the star symbol. These processes are denoted as MS1, MS2, MS3+O.
  • These processing might be, but not limited to, adaptive quantization, latent sample scaling, and latent sample offsetting operations.
  • an adaptive quantization process might correspond to scaling of a sample with multiplier before the prediction process, wherein the multiplier is predefined or whose value is indicated in the bitstream.
  • the latent scaling process might correspond to the process where a sample is scaled with a multiplier after the prediction process, wherein the value of the multiplier is either predefined or indicated in the bitstream.
  • the 19 F1250625PCT offsetting operation might correspond to adding an additive element to the sample, again wherein the value of the additive element might be indicated in the bitstream or inferred or predetermined.
  • Another operation might be tiling operation, wherein samples are first tiled (grouped) into overlapping or non-overlapping regions, wherein each region is processed independently. For example the samples corresponding to the luma component might be divided into tiles with a tile height of 20 samples, whereas the chroma components might be divided into tiles with a tile height of 10 samples for processing.
  • Another operation might be application of wavefront parallel processing. In wavefront parallel processing, a number of samples might be processed in parallel, and the amount of samples that can be processed in parallel might be indicated by a control parameter.
  • the said control parameter might be indicated in the bitstream, be inferred, or can be predetermined.
  • the number of samples that can be processed in parallel might be different, hence different indicators can be signalled in the bitstream to control the operation of luma and chrome processing separately.
  • 2.7 Colors separation and conditional coding In one example the primary and secondary color components of an image are coded separately, using networks with similar architecture, but different number of channels as shown in 8. All boxes with same names are sub-networks with the similar architecture, only input-output tensor size and number of channels are different. Number of channels for primary component is ⁇ ⁇ 128, for secondary components is ⁇ ⁇ 64.
  • ⁇ ⁇ Prior to the encoding ⁇ ⁇ , ⁇ ⁇ goes through a module which adjusts the sample location by down- sampling (marked as “s ⁇ ” on Fig.1), this essentially means that coded picture size for secondary component is different from the coded picture size for primary component.
  • the scaling factor s is variable, but the default scaling factor is ⁇ ⁇ 2.
  • the size of auxiliary input tensor in conditional coding is adjusted in order the encoder receives primary and secondary components tensor with the same picture size.
  • the secondary component is rescaled to the original picture size with a neural-network based upsampling filter module (“NN-color filter s ⁇ ” on Fig.
  • Fig. 8 exemplifies an image coding system, where the input image is first transformed into primary (Y) and secondary components (UV). The outputs ⁇ ⁇ , ⁇ ⁇ are the reconstructed outputs corresponding to the primary and secondary components. At the and of the processing, ⁇ ⁇ , ⁇ ⁇ are converted back to RGB color format. Typically the ⁇ ⁇ is downsampled (resized) before processing with the encoding and decoding modules (neural networks). For example the size of the ⁇ ⁇ might be reduced by a factor of 50% in each of the vertical and horizontal dimensions.
  • Fig. 9 illustrates synthesis transform example for learning based image coding.
  • the example synthesis transform above includes a sequence of 4 convolutions with up-sampling with stride of 2.
  • the synthesis transform sub-Net is depicted on Fig.9.
  • the size of the tensor in different parts of synthesis transform before cropping layer is the diagram on Fig.9.
  • Tile size ⁇ ⁇ 0, ... , ⁇ ⁇ 1 , ⁇ ⁇ 0, ... , ⁇ ⁇ ⁇ 1 a ⁇ ⁇ ⁇ nd ⁇ ⁇ 0, ... , ⁇ ⁇ 1, ⁇ ⁇ ⁇ .
  • beta_displacement_log_plus_2048[comp] minus 2048 is a – parameter indicating a displacement between the rate control parameter beta selected by encoder for comp component and the reference rate control parameter beta associated with the index of the used model (model_id). The displacement is in logarithmic scale.
  • betaDisplacementLog[comp] beta_displacement_log_plus_2048[comp] – 2 11
  • the reference rate control parameter beta ( ⁇ ) is the parameter used in the model training to control the ratio between the bitrate and distortion.
  • the model here means the model associated with the index of the used model (model_id).
  • Color_transform_idx index indicating transformation between internal and output color format
  • color_transform_idx 0 indicates no color transformation
  • color_transform_idx 1 indicates color transformation as in ITU-R BT.709
  • color_transform_idx 2 indicates color transformation with user defined parameters.
  • color_transform_matrix[i][j] is a matrix of color convertion, signalled only for color_transform_idx equal to 2.
  • Tools information semantics tools_header_size is the number of bytes in the tools header excluding the first two-byte marker; rvs_enable_flag[comp] – is a flag used in the RVS for comp component.0 indicates RVS disabled, 1 indicates RVS turn on. rvs_num_idx[comp] is an identificator for operation point, 0 means 1 threshold for RVS, 1 means 3 thresholds for RVS. lsbs_enable_flag[comp] – is a flag used in the LSBS mode for comp component.0 indicates LSBS disabled, 1 indicates LSBS turn on.
  • grfs_enable_flag[comp] is an enable flag for channel-wise gain unit refinement scale tool, if grfs_enable_flag[comp] is equal to one then residual tensor elements in channel indicated by grfs_channel_flag[comp] for comp component are scalled.
  • grfs_channel_flag[comp] is an array with 1-bit flags. Its size depends on value of the comp. For comp equals to 0 the size is C p and for comp equals to 1 the size is C s .
  • the value of fl[0] is 40 F1250625PCT restricted to be smaller than 4 and greater than 0.
  • the value of fl[1] is restricted to be smaller than 4 and greater than 0.
  • nonLinear_enabled_U_flag – on/off switch for nonlinear filtering process of U component.
  • nonLinear_enabled_V_flag – on/off switch for nonlinear filtering process of V component.
  • nonlinearW – width of the weight tensor of the nonlinear filtering process.
  • NonlinerW shall be multiple of 64.
  • nonlinearH height of the weight tensor of the nonlinear filtering process.
  • q_primary_sigma_Idx is a 1D array of size h4 ⁇ w4 and each element in this array is equal to q_sigma_Idx[quality_map_entropy_index[comp]], where the q_sigma_Idx[k] is derived according to Table 4.
  • Hyper Scale Decoder Hyper scale decoder net is depicted in Fig.17.
  • the input of hyper scale decoder is: ⁇ ⁇ C, h 6 , w 6 ⁇ reconstructed hyper tensor after me-tANS decoder (described in sections 9.6), 45 F1250625PCT ⁇ model parameters for Hyper Scale Decoder Net defined by modelIdx, all multiplier parameters in those models are 8-bits integer.
  • hyper scale decoder is standard deviation logarithm tensor I ⁇ ⁇ C, h 6 , w 6 ⁇ with integer values in a range 0 ⁇ I ⁇ ⁇ (( N ⁇ ⁇ 1) ⁇ sigmaPrecision), where N ⁇ , sigmaPrecision are defined in section 15.4.
  • h 4 , w 4 , h 6 , w 6 are defined in the Table 1. All operation in scalable hyper decoder are integer, an accumulator in all computations is within 32 bits integer diapason, multipliers in model parameters are quantized to 8-bits integer. This guarantees bit-exact behaviour of this neural network module.
  • Hyper scale decoder uses special neural network layers quantized convolutions.
  • Hyper scale decoder starts with quantized convolution kernel size 1 ⁇ 1, stride 1, followed by rectified linear unit. Then there is a stride 1 quantized convolution with kernel size 3 ⁇ 3, followed by rectified linear unit. The next stride 1 quantized convolution (kernel size 1 ⁇ 1) increases the number of channels to 16 ⁇ C . It is followed by shuffle (stride 4), which brings number of channels back to ⁇ . The cropping layer (to size h4, w4) ensures the size of output tensor is ⁇ C, h4, w4 ⁇ . The process concluded with abs operation.
  • Model with weights are storied in electronic attachment to this document, in format specicied in section 4.8 and locates in model_ ⁇ MID>/ ⁇ COMP>/hyper_scale_decoder.onnx, where ⁇ MID> is an integer from 0 to 3, ⁇ COMP> is “primary” or “secondary”. 0.
  • the input of this process is: ⁇ ⁇ C, h6, w6 ⁇ reconstructed hyper tensor after me-tANS decoder (described in sections 46 F1250625PCT 9.6), ⁇ model parameters for Hyper Decoder Net defined by pair modelIdx.
  • the output of this process is: ⁇ ⁇ ⁇ ⁇ 4 ⁇ C, h5, w5 ⁇ is re-shuffled explicit prediction tensor(part of predicton tensor derived from explicitly signalled information).
  • Hyper decoder process is depicted in Fig.18. Hyper decoder starts stride 1 convolution with kernel size 1 ⁇ 1, followed by inverse convolution (stride 2, kernel size 4 ⁇ 4), cropping layer (depth 6) and leacky rectified linear unit.
  • Next step is stride 1 convolution with kernel size 3 ⁇ 3, followed by leacky rectified linear unit. Number of channels keeps un-changed till this point (equal to number of channels ⁇ of input tensor). Then a stride 1 convolution layer with kernel size 3 ⁇ 3 increases number of channels to 4 ⁇ , and followed by leacky rectified linear unit. Hyper decoder concluded by stride 1 convolution with kernel size 1 ⁇ 1 followed by leacky rectified linear unit.
  • Model with weights are storied in electronic attachment to this document, in format specicied in section 4.8 and locates in model_ ⁇ MID>/ ⁇ COMP>/hyper_decoder.onnx, where ⁇ MID> is an integer from 0 to 3, ⁇ COMP> is “primary” or “secondary”. 1.
  • Latent tensor reconstruction process The input of this process is: ⁇ ⁇ C, h4, w4 ⁇ reconstructed residual tensor, which is an output of SKIP Model process (13.3.4), ⁇ ⁇ 4 ⁇ C, h5, w5 ⁇ re-shuffled explicit prediction tensor, which is an output of Hyper Decoder (11.2), The output of this process is: ⁇ ⁇ ′ ⁇ C, h 4 , w 4 ⁇ reconstructed latent tensor. ⁇ is obtained from ⁇ and ⁇ though Multi-stage Context Modelling process (section 11.4). 2.
  • Models with weights are storied in electronic attachment to this document, in format specicied in section 4.8 and locates in model_ ⁇ MID>/ ⁇ COMP>/MCM/stage ⁇ N>.onnx, where ⁇ MID> is an integer from 0 to 3, ⁇ COMP> is “primary” or “secondary”, ⁇ N> is a number of MCM’s stage from 0 to 3. 3. (10.3)
  • Skip Mode allows skip writing to / parsing from the bit-stream residual tensor elements which can be identified by encoder and decoder to be zeros. 48 F1250625PCT 4.
  • the output of Cube Flag generation process is: – cube_flag_skip[ ⁇ ,h ⁇ , ⁇ ⁇ ].
  • the cube size is ⁇ ⁇ 16 ⁇ 16. Each cube contains 4 cube flags which store in the 1D arrays ⁇ _ ⁇ .
  • the output of SKIP Mask generation process is: – mask_skip[C,h4,w4].
  • the output of this process is: ⁇ enhanced secondary component ⁇ ′ ⁇ ⁇ 2, ⁇ , ⁇ ⁇ which goes to the ICCI filter block (section 14.1) and ⁇ ′′ ⁇ 2, ⁇ ⁇ , ⁇ ⁇ the non-linear filter block (section ⁇ ⁇ 14.3). If is equal to 1 the following ordered steps are performed: ⁇
  • the parsing process according to parsing table in section 9.4.1 is invoked to obtain ⁇ ⁇ , ⁇ ⁇ , ⁇ ⁇ , ⁇ ⁇ and ⁇ ⁇ .
  • ⁇ Tiling process is as described in section 14.2.2 is invoked with parsed syntax elements as inputs and Tile1 tensor as output.
  • Luma Edge Filtering (LEF) filter This section details the Luma Edge Filtering(LEF) process.
  • the input of this process is: – ⁇ ′ ⁇ 1, ⁇ , ⁇ ⁇ (the output of ICCI process 14.1); ⁇ ⁇ ⁇ – log domain standard deviation tensor ⁇ ⁇ ′ [C,h 4 ,w 4 ] for primary component after Sigma Scale (10.4).
  • the output of this process is ⁇ ⁇ ⁇ , ⁇ .
  • ⁇ Picture header parsing process is invoked as described in section 9.4to obtain ⁇ ; ⁇ ⁇ h ⁇ is set equal to ⁇ _ ⁇ h ⁇ _ ⁇ _ ⁇ ; ⁇ The ⁇ ⁇ _ ⁇ ⁇ 1, ⁇ , W ⁇ tensor is obtained as follows, for ⁇ ⁇ 0.. ⁇ ⁇ 1, ⁇ ⁇ 0..
  • ⁇ ⁇ ⁇ 1; ⁇ ⁇ ⁇ ⁇ 0, ⁇ , ⁇ ⁇ h ⁇ , ⁇ ⁇ ⁇ , ⁇ ⁇ ⁇ ; ⁇ ⁇ intensity[4] luma_edge_filter_intensity_list [targetBppIdx]; ⁇ ⁇ ′ ⁇ ⁇ ⁇ ′ ⁇ 255; ⁇ ⁇ ⁇ ⁇ ′ ⁇ ⁇ ⁇ ; ⁇ 8190; 51 F1250625PCT 0, ⁇ ⁇ ⁇ ⁇ ⁇ 0, ⁇ , ⁇ ⁇ ⁇ h ⁇ 0 ⁇ 1, ⁇ h ⁇ 0 ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ 0, ⁇ , ⁇ ⁇ ⁇ h ⁇ 1 ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ In this when the indices of a tensor exceeds the tensor boundaries.
  • beta_displacement_log_plus_2048[comp] minus 2048 is a – parameter indicating a displacement between the rate control parameter beta selected by encoder for comp component and the reference rate control parameter beta associated with the index of the used model (model_id). The displacement is in logarithmic scale.
  • betaDisplacementLog[comp] beta_displacement_log_plus_2048[comp] – 2 11
  • the reference rate control parameter beta ( ⁇ ) is the parameter used in the model training to control the ratio between the bitrate and distortion.
  • the model here means the model associated with the index of the used model (model_id).
  • betaDisplacementLog[1] betaDisplacementLog[0].
  • Each model_id is associated with the reference r ate control parameter ⁇ , which was used in the model training to control the ratio between the bitrate and distortion.
  • the forward gain tensor m is used at encoder side in Gain Unit.
  • Forward gain tensor in logarithmic scale ⁇ ⁇ is used in Sigma scale.
  • the inverse gain tensor ⁇ ⁇ is used at decoder side in Inverse Gain Unit. All three forward ⁇ , inverse ⁇ ⁇ and logarithmics domain ⁇ ⁇ gain tensors have size ⁇ , h ⁇ , ⁇ equal to the size of the residual tensor for the component.
  • Component index comp is equal to 0 for the primary color component and 1 for the secondary component.
  • the input of gain tensor derivation process is: ⁇ 12-bits signed variable betaDisplacementLog[comp]; 53 F1250625PCT ⁇ tensor with 3-bit elements ⁇ _ ⁇ _ ⁇ h ⁇ , ⁇ ⁇ ⁇ ; ⁇ index model_id which indicates pretrained model to be used.
  • the output of gain tensor derivation process is: ⁇ forward gain tensor in logarithmic scale ⁇ ⁇ , h ⁇ , ⁇ ; ⁇ forward gain tensor ⁇ ⁇ , h ⁇ , ⁇ ; ⁇ inverse gain tensor ⁇ ⁇ , h ⁇ , ⁇ . Sizes of all tensors ⁇ , h ⁇ , ⁇ for primary and secondary components are defined in Table 1.
  • Forward gain tensor in logarithmic scale ⁇ ⁇ is computed as following: For 0 ⁇ ⁇ ⁇ ⁇ ⁇ ; 0 ⁇ ⁇ ⁇ h ⁇ ; 0 ⁇ ⁇ ⁇ ⁇ : ⁇ , ⁇ , ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ 3 ⁇ _ ⁇ , ⁇ ⁇ where quality_map[i,j] is equal to 0 when gain_3D_enable_flag[comp] is equal to 0 and, otherwise, is equal to ⁇ 8,8, ⁇ _ ⁇ _ ⁇ , ⁇ ⁇ ⁇ _ ⁇ _ ⁇ comp ⁇ ⁇ , ⁇ , ⁇ 16 ⁇ ⁇ _ ⁇ _ ⁇ ⁇ , ⁇ ⁇ 16 o ⁇ _ ⁇ _ ⁇ , ⁇ ⁇ ⁇ ⁇ 0.
  • the input of this process are: ⁇ forward gain tensor in logarithmic scale m log , ⁇ standard deviation logarithm tensor I ⁇ of size ⁇ C, h 4 , w 4 ⁇ , which is and output of Hyper Scale Decoder.
  • the output of this process is: ⁇ scaled by forward gain tensor standard deviation logarithm tensor I’ ⁇ of size ⁇ C, h 4 , w 4 ⁇ which goes to the adaptive sigma scale.
  • Fig.22 illustrates an MCM model to obtain cube_flag, which will be encoded to the bitstreams.
  • the ⁇ _ ⁇ denotes the latent tensor after analysis transform, and ⁇ _ ⁇ is the final dequantized residual.
  • the input ⁇ _ ⁇ will 56 F1250625PCT first down-shuffled into 4 sub-tensors, whose spatial size (width or height) is half of the input.
  • variable rate support In image and video coding, it is sometimes necessary to generate a bitstream that is close to a target bitrate (e.g. variable rate coding). However, achieving this requires additional processing, and hence increases the complexity. In majority of the applications, rate matching (variable rate support) is not used, as the application has a target quality, and not a target bitrate. Therefore, the supporting variable rate coding increases the burden on majority of the decoders without any benefit.
  • one common method is to scale parameters that are encoded in the bitstream before quantization, and applying an inverse scaling after quantization.
  • variable rate support as described in the above section 2.15.
  • series of operations need to be performed by all decoders, although the variable rate coding is used only by certain applications. 4.
  • the solutions should be considered as examples to explain the general concepts and should not be interpreted in a narrow way. Furthermore, these solutions can be applied individually or combined in any manner.
  • the proposed solution(s) primarily includes three aspects: a) skip mode signaling; 2) bitstream structure; 3) variable rate support.
  • 4.1 Skip mode tool header signaling In the above Section 8 (6.3.1.1) Model header, skip_mode_header() is described, wherein cube_flag for the whole image is encoded to the bistreams. In this solution, it is proposed to reorganize the cube_flag.
  • the image is divided into regions, thereby, multiple region_flag tensors will be derived from these regions, forming a cube_flag list.
  • N NumHorSplits * NumVerSplits.
  • the encoder will encode these N tensors sequentially. 1) In one example, cube_flag tensor is reorganized, wherein each region produces a distinct cube_flag.
  • the encoder encodes ⁇ cube_flag tensors sequentially, where ⁇ indicates the number of regions. 2) Alternatively, in one example, a list of cube_flag_header(), for example, cube_flag_list_header() is added for skip mode.
  • a list of cube_flag_header() for example, cube_flag_list_header() is added for skip mode.
  • the following table shows modified skip mode signaling, wherein encoding cube_flag_header() is repeated for N times, where N denotes the number of regions. Most relevant parts that have been added or modified are shown by using bolded words (e.g., this format indicates added text), and some of the deleted parts are shown by using words in italics between double curly brackets (e.g., ⁇ this format indicates deleted text ⁇ ).
  • Fig.23 illustrates the bitstream structure of q-stream, RY-stream and Ruv-stream are split based on regions, while z-stream as a single without split based on regions. As detailed in Fig. 23, it is proposed that to split q-stream, RY-stream and RUV-stream in the bitstreams based on the regions, while keep z-stream as a single stream.
  • SKIP Mask generation process is: – mask_skip[C, h4, w4].
  • the output of Cube Flag generation process is: – cube_flag_skip_list[numH
  • the cube size is ⁇ ⁇ 8 ⁇ 8
  • the 1D arrays ⁇ _ ⁇ is represent a 2D array with height h and width ⁇ in raster scan order (an example is shown in Fig. 27), where h ⁇ ⁇ h ⁇ ⁇ 7 ⁇ ⁇ 3, ⁇ ⁇ ⁇ ⁇ 7 ⁇ ⁇ 3.
  • variable rate support 5.5.1 Decoder/encoder solution 1 Performing a conversion using a neural network between a bitstream and a reconstructed image or video based on an indicator in the bitstream, wherein, ⁇ If the indicator indicates a positive action, variable rate coding is used. ⁇ Otherwise variable rate coding is not used. 5.5.2 Decoder/encoder solution 2 Performing a conversion using a neural network between a bitstream and a reconstructed image or video based on an indicator in the bitstream, wherein, ⁇ If the indicator indicates a positive action, a value obtained from a bitstream is used in variable rate coding process. ⁇ Otherwise a default value is used in variable rate coding process.
  • Decoder/encoder solution 3 Performing a conversion using a neural network between a bitstream and a reconstructed image or video based on an indicator in the bitstream, wherein, ⁇ If the indicator indicates a positive action, a value obtained from a bitstream is used in variable rate coding process. 66 F1250625PCT ⁇ Otherwise a default value from a set of default values is used in variable rate coding process. 5.5.4 According to the above solution 3, the default value is selected from a set of default values based on an index indicated in the bitstream. 6.
  • Model header model_header(comp) Descriptor v ri bl r t din n bl d[ m ]
  • variable_rate_coding_enabled[comp] is a flag (false/true), indicating if variable rate coding is enabled or disabled.
  • variable_rate_coding_enabled[comp] is equal to false
  • beta_displacement_log_ plus_2048[comp] is inferred to be equal to 2048.
  • independent_beta_uv is a flag (false/true)
  • independent_beta_uv indicates that the beta displacement parameter for primary and secondary components are different. If independent_beta_uv equal to false then beta displacement parameter for primary and secondary components are the same.
  • beta_displacement_log_plus_2048[comp] minus 2048 is a – parameter indicating a displacement between the rate control parameter beta selected by encoder for comp component and the reference rate control parameter beta associated with the index of the used model (model_id).
  • beta_displacement_log_plus_2048[comp] minus 2048 is a – parameter indicating a displacement between the rate control parameter beta selected by encoder for comp component and the reference rate control parameter beta associated with the index of the used model (model_id). The displacement is in logarithmic scale.
  • betaDisplacementLog[comp] beta_displacement_log_plus_2048[comp] – 2 11
  • the reference rate control parameter beta ( ⁇ ) is the parameter used in the model training to control the ratio between the bitrate and distortion.
  • the model here means the model associated with the index of the used model (model_id).
  • betaDisplacementLog[1] betaDisplacementLog[0].
  • an NN-based model may be a model based on neural network technologies.
  • an NN-based model may specify sequence of neural network modules (also called architecture) and model parameters.
  • the neural network module may comprise a set of neural network layers. Each neural network layer specifies a tensor operation which receives and outputs tensor, and each layer has trainable parameters. It should be understood that the possible implementations of the NN-based model described here are merely illustrative and therefore should not be construed as limiting the present disclosure in any way.
  • the codestream comprises at least one indication associated with a usage of a variable rate coding process in the conversion.
  • the at least one indication may comprise a first indication indicating whether the variable rate coding process is enabled or disabled. In this case, it is possible to turn on or turn off the variable rate functionality. Thereby, it is possible to skip operations only necessary for performing variable rate coding process in a case where the variable rate coding process is disabled, and thus the coding efficiency can be improved.
  • the at least one indication may also comprise any other suitable indication(s), a part of which will be described in detail below.
  • the at least one indication may comprise a second indication indicating whether a first parameter or a second parameter is used in the variable rate coding process.
  • the first parameter is predetermined, and the second parameter is obtained based on the codestream.
  • the first parameter and/or the second parameter may be used to adjust a residual parameter or a standard deviation parameter (a.k.a., sigma parameter) for the visual data.
  • the second parameter may be determined as follows: ⁇ ⁇ , ⁇ , ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ where ⁇ ⁇ ⁇ represents a logarithmic domain gain tensor, ⁇ represents a channel index, ⁇ and ⁇ represent two spatial indexes, ⁇ ⁇ represents a model index, ⁇ represents a color component index, ⁇ represents a logarithmic domain displacement between the rate control parameter beta selected by encoder for a color component and the reference rate control parameter beta associated with the index of the used model ( ⁇ ).
  • ⁇ , ⁇ , ⁇ may correspond to the second parameter
  • ⁇ may correspond to the first parameter
  • ⁇ may be indicated in the codestream.
  • the first parameter in accordance with a determination that the second indication is equal to a first value (such as 0, FALSE, or the like), the first parameter may be used in the variable rate coding process. In accordance with a determination that the second indication is equal to a second value (such as 1, TRUE, or the like), the second parameter may be used in the variable rate coding process.
  • the at least one indication may comprise a third indication indicating whether one or more indications associated with the variable rate coding process are comprised in the codestream.
  • the at least one indication may comprise a fourth indication indicating whether a value of a third parameter associated with the variable rate coding process is inferred to be a predetermined value.
  • the predetermined value may be 0 or 2048.
  • the third parameter may be a displacement between a rate control parameter selected by an encoder and a reference rate control parameter associated with an index of the used model.
  • the third parameter may be the displacement plus 2048.
  • the at least one indication may be comprised in a header syntax structure in the codestream.
  • the header syntax structure may be a picture header. In the picture header, a model header immediately follows the at least one indication.
  • the header syntax structure may be a model header.
  • the at least one indication may be at the initial position within the model header.
  • a non-transitory computer-readable recording medium stores a codestream of visual data which is generated by a method performed by an apparatus for visual data processing. The method comprises: performing a conversion from the visual data to the codestream with a neural network (NN)-based model, wherein the codestream comprises at least one indication associated with a usage of a variable rate coding process in the conversion.
  • NN neural network
  • a method for visual data processing comprising: performing a conversion between visual data and a codestream of the visual data with a neural network (NN)-based model, wherein the codestream comprises at least one indication associated with a usage of a variable rate coding process in the conversion.
  • the at least one indication comprises a first indication indicating whether the variable rate coding process is enabled or disabled.
  • Clause 3. The method of any of clauses 1-2, wherein the at least one indication comprises a second indication indicating whether a first parameter or a second parameter is used in the variable rate coding process, the first parameter being predetermined and the second parameter being obtained based on the codestream.
  • Clause 5 The method of any of clauses 3-4, wherein the first parameter is determined from a plurality of predetermined parameters based on an index.
  • Clause 6. The method of clause 5, wherein the index comprises a model index, or the index is indicated in the codestream.
  • Clause 7. The method of any of clauses 3-6, wherein in accordance with a determination that the second indication is equal to a first value, the first parameter is used in the variable rate coding process, or in accordance with a determination that the second indication is equal to a second value, the second parameter is used in the variable rate coding process.
  • Clause 8 The method of any of clauses 3-7, wherein the second parameter is determined based on a sum of the first parameter and a value indicated in the codestream.
  • Clause 9. The method of any of clauses 1-8, wherein the at least one indication comprises a third indication indicating whether one or more indications associated with the variable rate coding process are comprised in the codestream.
  • Clause 10. The method of clause 9, wherein the one or more indications comprise at least one of the following: an indication indicating whether a rate control parameter for a primary component and a rate control parameter for a secondary component are different, or an indication indicating a displacement between a rate control parameter selected by an encoder and a reference rate control parameter associated with an index of the used model.
  • Clause 11 The method of any of clauses 9-10, wherein the third indication is 73 F1250625PCT represented as beta_displacement_present_flag, beta_displacement_inferred_flag, or variable_rate_coding_enabled flag.
  • Clause 12 The method of any of clauses 1-11,wherein the at least one indication comprises a fourth indication indicating whether a value of a third parameter associated with the variable rate coding process is inferred to be a predetermined value.
  • Clause 13 The method of clause 12, wherein the predetermined value is 0 or 2048.
  • Clause 15 The method of any of clauses 1-14, wherein the at least one indication is comprised in a header syntax structure in the codestream.
  • Clause 16 The method of clause 15, wherein the header syntax structure is a picture header.
  • Clause 17 The method of clause 16, wherein in the picture header, a model header immediately follows the at least one indication.
  • Clause 18 The method of clause 15, wherein the header syntax structure is a model header. [0089] Clause 19.
  • Clause 24 A non-transitory computer-readable storage medium storing instructions that cause a processor to perform a method in accordance with any of clauses 1-22.
  • Clause 25 A non-transitory computer-readable recording medium storing a codestream of visual data which is generated by a method performed by an apparatus for visual data processing, wherein the method comprises: performing a conversion from the 74 F1250625PCT visual data to the codestream with a neural network (NN)-based model, wherein the codestream comprises at least one indication associated with a usage of a variable rate coding process in the conversion.
  • NN neural network
  • the computing device 2900 includes a general-purpose computing device 2900.
  • the computing device 2900 may at least comprise one or more processors or processing units 2910, a memory 2920, a storage unit 2930, one or more communication units 2940, one or more input devices 2950, and one or more output devices 2960.
  • the computing device 2900 may be implemented as any user terminal or server terminal having the computing capability.
  • the server terminal may be a server, a large-scale computing device or the like that is provided by a service provider.
  • the user terminal may for example be any type of mobile terminal, fixed terminal, or portable terminal, including a mobile phone, station, unit, device, multimedia computer, multimedia tablet, Internet node, communicator, desktop computer, laptop computer, notebook computer, netbook computer, tablet computer, personal communication system (PCS) device, personal navigation device, personal digital assistant (PDA), audio/video player, digital camera/video camera, positioning device, television receiver, radio broadcast receiver, E-book device, gaming device, or any combination thereof, including the accessories and peripherals of these devices, or any combination thereof.
  • the computing device 2900 can support any type of interface to a user (such as “wearable” circuitry and the like).
  • the memory 2920 can be a volatile memory (for example, a register, cache, Random Access Memory (RAM)), a non-volatile memory (such as a Read-Only Memory (ROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), or a flash memory), or any combination thereof.
  • the storage unit 2930 may be any detachable or non-detachable medium and may include a machine-readable medium such as a memory, flash memory drive, magnetic disk or another other media, which can be used for storing information and/or visual data and can be accessed in the computing device 2900.
  • the computing device 2900 may further include additional detachable/non- detachable, volatile/non-volatile memory medium.
  • the input device 2950 may be one or more of a variety of input devices, such as a mouse, keyboard, tracking ball, voice-input device, and the like.
  • the output device 2960 may be one or more of a variety of output devices, such as a display, loudspeaker, printer, and the like.
  • the computing device 2900 can further communicate with one or more external devices (not shown) such as the storage devices and display device, with one or more devices enabling the user to interact with the computing device 2900, or any devices (such as a network card, a modem and the like) enabling the computing device 2900 to communicate with one or more other computing 76 F1250625PCT devices, if required.
  • cloud computing instead of being integrated in a single device, some or all components of the computing device 2900 may also be arranged in cloud computing architecture.
  • the components may be provided remotely and work together to implement the functionalities described in the present disclosure.
  • cloud computing provides computing, software, visual data access and storage service, which will not require end users to be aware of the physical locations or configurations of the systems or hardware providing these services.
  • the cloud computing provides the services via a wide area network (such as Internet) using suitable protocols.
  • a cloud computing provider provides applications over the wide area network, which can be accessed through a web browser or any other computing components.
  • the software or components of the cloud computing architecture and corresponding visual data may be stored on a server at a remote position.
  • the computing resources in the cloud computing environment may be merged or distributed at locations in a remote visual data center.
  • Cloud computing infrastructures may provide the services through a shared visual data center, though they behave as a single access point for the users. Therefore, the cloud computing architectures may be used to provide the components and functionalities described herein from a service provider at a remote location. Alternatively, they may be provided from a conventional server or installed directly or otherwise on a client device.
  • the computing device 2900 may be used to implement visual data encoding/decoding in embodiments of the present disclosure.
  • the memory 2920 may include one or more visual data coding modules 2925 having one or more program instructions.
  • the input device 2950 may receive visual data as an input 2970 to be encoded.
  • the visual data may be processed, for example, by the visual data coding module 2925, to generate an encoded bitstream.
  • the encoded bitstream may be provided via the output device 2960 as an output 2980.
  • the input device 2950 may receive an encoded bitstream as the input 2970.
  • the encoded bitstream may be processed, for example, by the visual data coding module 2925, to generate decoded visual data.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Signal Processing (AREA)
  • Multimedia (AREA)
  • Neurology (AREA)
  • Probability & Statistics with Applications (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)

Abstract

Embodiments of the present disclosure provide a solution for visual data processing. A method for visual data processing is proposed. The method comprises: performing a conversion between visual data and a codestream of the visual data with a neural network (NN)-based model, wherein the codestream comprises at least one indication associated with a usage of a variable rate coding process in the conversion.

Description

METHOD, APPARATUS, AND MEDIUM FOR VISUAL DATA PROCESSING FIELDS [0001] Embodiments of the present disclosure relates generally to visual data processing techniques, and more particularly, to neural network-based visual data coding. BACKGROUND [0002] The past decade has witnessed the rapid development of deep learning in a variety of areas, especially in computer vision and image processing. Neural network was invented originally with the interdisciplinary research of neuroscience and mathematics. It has shown strong capabilities in the context of non-linear transform and classification. Neural network- based image/video compression technology has gained significant progress during the past half decade. It is reported that the latest neural network-based image compression algorithm achieves comparable rate-distortion (R-D) performance with Versatile Video Coding (VVC). With the performance of neural image compression continually being improved, neural network-based video compression has become an actively developing research area. However, coding flexibility of neural network-based image/video coding is generally expected to be further improved. SUMMARY [0003] Embodiments of the present disclosure provide a solution for visual data processing. [0004] In a first aspect, a method for visual data processing is proposed. The method comprises: performing a conversion between visual data and a codestream of the visual data with a neural network (NN)-based model, wherein the codestream comprises at least one indication associated with a usage of a variable rate coding process in the conversion. [0005] Based on the method in accordance with the first aspect of the present disclosure, at least one indication associated with a usage of a variable rate coding process in the conversion is signaled in the bitstream. Compared with the conventional solution, the proposed method can better support the application of variable rate coding process. Thereby, the coding flexibility can be improved. [0006] In a second aspect, an apparatus for visual data processing is proposed. The apparatus comprises a processor and a non-transitory memory with instructions thereon. The instructions upon execution by the processor, cause the processor to perform a method in accordance with the first aspect of the present disclosure. [0007] In a third aspect, a non-transitory computer-readable storage medium is proposed. The non-transitory computer-readable storage medium stores instructions that cause a 1 F1250625PCT processor to perform a method in accordance with the first aspect of the present disclosure. [0008] In a fourth aspect, another non-transitory computer-readable recording medium is proposed. The non-transitory computer-readable recording medium stores a codestream of visual data which is generated by a method performed by an apparatus for visual data processing. The method comprises: performing a conversion from the visual data to the codestream with a neural network (NN)-based model, wherein the codestream comprises at least one indication associated with a usage of a variable rate coding process in the conversion. [0009] In a fifth aspect, a method for storing a codestream of visual data is proposed. The method comprises: performing a conversion from the visual data to the codestream with a neural network (NN)-based model; and storing the codestream in a non-transitory computer- readable recording medium, wherein the codestream comprises at least one indication associated with a usage of a variable rate coding process in the conversion. [0010] This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter. BRIEF DESCRIPTION OF THE DRAWINGS [0011] Through the following detailed description with reference to the accompanying drawings, the above and other objectives, features, and advantages of example embodiments of the present disclosure will become more apparent. In the example embodiments of the present disclosure, the same reference numerals usually refer to the same components. [0012] Fig. 1A illustrates a block diagram that illustrates an example visual data coding system in accordance with some embodiments of the present disclosure; [0013] Fig. 1B is a schematic diagram illustrating an example transform coding scheme; [0014] Fig. 2 illustrates example latent representations of an image; [0015] Fig. 3 is a schematic diagram illustrating an example autoencoder implementing a hyperprior model; [0016] Fig. 4 is a schematic diagram illustrating an example combined model configured to jointly optimize a context model along with a hyperprior and the autoencoder; [0017] Fig. 5 illustrates an example encoding process; [0018] Fig. 6 illustrates an example decoding process; [0019] Fig. 7 illustrates an example decoding process according to some embodiments of the present disclosure; [0020] Fig. 8 illustrates an example learning-based image codec architecture; 2 F1250625PCT [0021] Fig. 9 illustrates an example synthesis transform for learning based image coding; [0022] Fig. 10 illustrates an example leaky Rectified Linear Unit (ReLU) activation function; [0023] Fig. 11 illustrates an example ReLU activation function; [0024] Fig. 12 illustrates an example down-shuffle operation; [0025] Fig. 13 illustrates an example up-shuffle operation; [0026] Fig. 14 illustrates latent tiles in synthesis transform; [0027] Fig. 15 illustrates a bitstream layout; [0028] Fig. 16 illustrates an example decoder structure; [0029] Fig. 17 illustrates an example hyper scale decoder; [0030] Fig. 18 illustrates an example hyper decoder; [0031] Fig. 19 illustrates a diagram of an example multistage context modelling (MCM) structure; [0032] Fig. 20 illustrates an example implementation of primary component guided adaptive up-sampling filter; [0033] Fig. 21 illustrates a luma edge filtering (LEF) general process; [0034] Fig. 22 illustrates an MCM model to obtain cube_flag; [0035] Fig. 23 illustrates a bitstream structure in accordance with embodiments of the present disclosure; [0036] Fig. 24 illustrates another bitstream structure in accordance with embodiments of the present disclosure; [0037] Fig. 25 illustrates a further bitstream structure in accordance with embodiments of the present disclosure; [0038] Fig.26 illustrates a still further bitstream structure in accordance with embodiments of the present disclosure; [0039] Fig. 27 illustrates an example of reshaping a one-dimensional (1D) array of cube_flag into a two dimensional (2D) array; [0040] Fig. 28 illustrates a flowchart of a method for visual data processing in accordance with embodiments of the present disclosure; and [0041] Fig. 29 illustrates a block diagram of a computing device in which various embodiments of the present disclosure can be implemented. [0042] Throughout the drawings, the same or similar reference numerals usually refer to the same or similar elements. DETAILED DESCRIPTION [0043] Principle of the present disclosure will now be described with reference to some 3 F1250625PCT embodiments. It is to be understood that these embodiments are described only for the purpose of illustration and help those skilled in the art to understand and implement the present disclosure, without suggesting any limitation as to the scope of the disclosure. The disclosure described herein can be implemented in various manners other than the ones described below. [0044] In the following description and claims, unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skills in the art to which this disclosure belongs. [0045] References in the present disclosure to “one embodiment,” “an embodiment,” “an example embodiment,” and the like indicate that the embodiment described may include a particular feature, structure, or characteristic, but it is not necessary that every embodiment includes the particular feature, structure, or characteristic. Moreover, such phrases are not necessarily referring to the same embodiment. Further, when a particular feature, structure, or characteristic is described in connection with an example embodiment, it is submitted that it is within the knowledge of one skilled in the art to affect such feature, structure, or characteristic in connection with other embodiments whether or not explicitly described. [0046] It shall be understood that although the terms “first” and “second” etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first element could be termed a second element, and similarly, a second element could be termed a first element, without departing from the scope of example embodiments. As used herein, the term “and/or” includes any and all combinations of one or more of the listed terms. [0047] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments. As used herein, the singular forms “a”, “an” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms “comprises”, “comprising”, “has”, “having”, “includes” and/or “including”, when used herein, specify the presence of stated features, elements, and/or components etc., but do not preclude the presence or addition of one or more other features, elements, components and/ or combinations thereof. Example Environment [0048] Fig. 1A is a block diagram that illustrates an example visual data coding system 100 that may utilize the techniques of this disclosure. As shown, the visual data coding system 100 may include a source device 110 and a destination device 120. The source device 110 can be also referred to as a visual data encoding device, and the destination device 120 4 F1250625PCT can be also referred to as a visual data decoding device. In operation, the source device 110 can be configured to generate encoded visual data and the destination device 120 can be configured to decode the encoded visual data generated by the source device 110. The source device 110 may include a visual data source 112, a visual data encoder 114, and an input/output (I/O) interface 116. [0049] The visual data source 112 may include a source such as a visual data capture device. Examples of the visual data capture device include, but are not limited to, an interface to receive visual data from a visual data provider, a computer graphics system for generating visual data, and/or a combination thereof. [0050] The visual data may comprise one or more pictures of a video or one or more images. The visual data encoder 114 encodes the visual data from the visual data source 112 to generate a bitstream. The bitstream may include a sequence of bits that form a coded representation of the visual data. The bitstream may include coded pictures and associated visual data. The coded picture is a coded representation of a picture. The associated visual data may include sequence parameter sets, picture parameter sets, and other syntax structures. The I/O interface 116 may include a modulator/demodulator and/or a transmitter. The encoded visual data may be transmitted directly to destination device 120 via the I/O interface 116 through the network 130A. The encoded visual data may also be stored onto a storage medium/server 130B for access by destination device 120. [0051] The destination device 120 may include an I/O interface 126, a visual data decoder 124, and a display device 122. The I/O interface 126 may include a receiver and/or a modem. The I/O interface 126 may acquire encoded visual data from the source device 110 or the storage medium/server 130B. The visual data decoder 124 may decode the encoded visual data. The display device 122 may display the decoded visual data to a user. The display device 122 may be integrated with the destination device 120, or may be external to the destination device 120 which is configured to interface with an external display device. [0052] The visual data encoder 114 and the visual data decoder 124 may operate according to a visual data coding standard, such as video coding standard or still picture coding standard and other current and/or further standards. [0053] Some example embodiments of the present disclosure will be described in detailed hereinafter. It should be understood that section headings are used in the present document to facilitate ease of understanding and do not limit the embodiments disclosed in a section to only that section. Furthermore, while certain embodiments are described with reference to Versatile Video Coding or other specific visual data codecs, the disclosed techniques are applicable to other coding technologies also. Furthermore, while some embodiments 5 F1250625PCT describe coding steps in detail, it will be understood that corresponding steps decoding that undo the coding will be implemented by a decoder. Furthermore, the term visual data processing encompasses visual data coding or compression, visual data decoding or decompression and visual data transcoding in which visual data are represented from one compressed format into another compressed format or at a different compressed bitrate. 1. Brief Summary The present disclosure is related to neural network (NN)-based image and video coding. Specifically, it is related to the method of signaling tool header of skip mode in support of regional accessibility, wherein regional accessibility refers to the capability of correctly decoding only a regional part of an image (also referred to as a picture) or a video. In addition, this disclosure is related to a neural network-based image and video compression method comprising modification of components of an image using convolution layers. The weights of the convolution layer is included in the bitstream. The ideas may be applied individually or in various combinations, for image and/or video coding methods and specifications. 2. Introduction The past decade has witnessed the rapid development of deep learning in a variety of areas, especially in computer vision and image processing. Inspired from the great success of deep learning technology to computer vision areas, many researchers have shifted their attention from conventional image/video compression techniques to neural image/video compression technologies. Neural network was invented originally with the interdisciplinary research of neuroscience and mathematics. It has shown strong capabilities in the context of non-linear transform and classification. Neural network-based image/video compression technology has gained significant progress during the past half decade. It is reported that the latest neural network- based image compression algorithm achieves comparable R-D performance with Versatile Video Coding (VVC), the latest video coding standard developed by Joint Video Experts Team (JVET) with experts from MPEG and VCEG. With the performance of neural image compression continually being improved, neural network-based video compression has become an actively developing research area. However, neural network-based video coding still remains in its infancy due to the inherent difficulty of the problem. 2.1 Image/video compression Image/video compression (also referred to as image/video coding) usually refers to the computing technology that compresses image/video into binary code to facilitate storage and transmission. The binary codes may or may not support losslessly reconstructing the original image/video, termed lossless compression and lossy compression. Most of the efforts are devoted to lossy compression since lossless reconstruction is not necessary in most scenarios. Usually the 6 F1250625PCT performance of image/video compression algorithms is evaluated from two aspects, i.e. compression ratio and reconstruction quality. Compression ratio is directly related to the number of binary codes, the less the better; Reconstruction quality is measured by comparing the reconstructed image/video with the original image/video, the higher the better. Image/video compression techniques can be divided into two branches, the classical video coding methods and the neural-network-based video compression methods. Classical video coding schemes adopt transform-based solutions, in which researchers have exploited statistical dependency in the latent variables (e.g., DCT or wavelet coefficients) by carefully hand- engineering entropy codes modeling the dependencies in the quantized regime. Neural network- based video compression is in two flavors, neural network-based coding tools and end-to-end neural network-based video compression. The former is embedded into existing classical video codecs as coding tools and only serves as part of the framework, while the latter is a separate framework developed based on neural networks without depending on classical video codecs. In the last three decades, a series of classical video coding standards have been developed to accommodate the increasing visual content. The international standardization organizations ISO/IEC has two expert groups namely Joint Photographic Experts Group (JPEG) and Moving Picture Experts Group (MPEG), and ITU-T also has its own Video Coding Experts Group (VCEG) which is for standardization of image/video coding technology. The influential video coding standards published by these organizations include JPEG, JPEG 2000, H.262, H.264/AVC and H.265/HEVC. After H.265/HEVC, the Joint Video Experts Team (JVET) formed by MPEG and VCEG has been working on a new video coding standard Versatile Video Coding (VVC). The first version of VVC was released in July 2020. An average of 50% bitrate reduction is reported by VVC under the same visual quality compared with HEVC. Neural network-based image/video compression is not a new invention since there were a number of researchers working on neural network-based image coding. But the network architectures were relatively shallow, and the performance was not satisfactory. Benefit from the abundance of data and the support of powerful computing resources, neural network-based methods are better exploited in a variety of applications. At present, neural network-based image/video compression has shown promising improvements, confirmed its feasibility. Nevertheless, this technology is still far from mature and a lot of challenges need to be addressed. 2.2 Neural networks Neural networks, also known as artificial neural networks (ANN), are the computational models used in machine learning technology which are usually composed of multiple processing layers and each layer is composed of multiple simple but non-linear basic computational units. One benefit of such deep networks is believed to be the capacity for processing data with multiple 7 F1250625PCT levels of abstraction and converting data into different kinds of representations. Note that these representations are not manually designed; instead, the deep network including the processing layers is learned from massive data using a general machine learning procedure. Deep learning eliminates the necessity of handcrafted representations, and thus is regarded useful especially for processing natively unstructured data, such as acoustic and visual signal, whilst processing such data has been a longstanding difficulty in the artificial intelligence field. 2.3 Neural networks for image compression Existing neural networks for image compression methods can be classified in two categories, i.e., pixel probability modeling and auto-encoder. The former one belongs to the predictive coding strategy, while the latter one is the transform-based solution. Sometimes, these two methods are combined together in literature. 2.3.1 Pixel probability modeling According to Shannon’s information theory, the optimal method for lossless coding can reach the minimal coding rate െ logଶ ^^^^^^ where ^^^^^^ is the probability of symbol ^^. A number of lossless coding methods were developed in literature and among them arithmetic coding is believed to be among the optimal ones. Given a probability distribution ^^^^^^, arithmetic coding ensures that the coding rate to be as close as possible to its theoretical limit െ logଶ ^^^^^^ without considering the rounding error. Therefore, the remaining problem is to how to determine the probability, which is however very challenging for natural image/video due to the curse of dimensionality. Following the predictive coding strategy, one way to model ^^^^^^ is to predict pixel probabilities one by one in a raster scan order based on previous observations, where ^^ is an image. ^^^^^^ ൌ ^^^^^^^^^^^^ଶ|^^^^…^^^^^^|^^^, … , ^^^ି^^…^^^^^^ൈ^|^^^, … , ^^^ൈ^ି^^ (1) where is also known as the context of the current pixel. When the image is large, it can be difficult to estimate the conditional probability, thereby a simplified method is to limit the range of its context. ^^^^^^ ൌ ^^^^^^^^^^^^ଶ|^^^^…^^^^^^|^^^ି^, … , ^^^ି^^…^^^^^^ൈ^|^^^ൈ^ି^, … , ^^^ൈ^ି^^ (2) where ^^ is a pre-defined constant controlling the range of the context. It should be noted that the condition may also take the sample values of other color components into consideration. For example, when coding the RGB color component, R sample is dependent on previously coded pixels (including R/G/B samples), the current G sample may be coded according to previously coded pixels and the current R sample, while for coding the current B sample, the previously coded pixels and the current R and G samples may also be taken into consideration. Neural networks were originally introduced for computer vision tasks and have been proven to be effective in regression and classification problems. Therefore, it has been proposed using neural 8 F1250625PCT networks to estimate the probability of ^^^^^^^ given its context ^^^, ^^ଶ, … , ^^^ି^. Most of the methods directly model the probability distribution in the pixel domain. Some researchers also attempt to model the probability distribution as a conditional one upon explicit or latent representations. That being said, it may be estimated that ^^^^^|^^^ ൌ ∏ ^ൈ^ ^ୀ^ ^^^^^^|^^^, … , ^^^ି^,^^^ (3) where ^^ is the additional condition and ^^^^^^ the is into an unconditional one and a information or high-level representations. 2.3.2 Auto-encoder Auto-encoder originates from the well-known work proposed by Hinton and Salakhutdinov. The method is trained for dimensionality reduction and consists of two parts: encoding and decoding. The encoding part converts the high-dimension input signal to low-dimension representations, typically with reduced spatial size but a greater number of channels. The decoding part attempts to recover the high-dimension input from the low-dimension representation. Auto-encoder enables automated learning of representations and eliminates the need of hand-crafted features, which is also believed to be one of the most important advantages of neural networks. Fig. 1B illustrates a typical transform coding scheme. The original image x is transformed by the analysis network gୟ to achieve the latent representation y. The latent representation y is quantized and compressed into bits. The number of bits R is used to measure the coding rate. The quantized latent representation y^ is then inversely transformed by a synthesis network g^ to obtain the reconstructed image x^. The distortion is calculated in a perceptual space by transforming x and x^ with the function g୮. It is intuitive to apply auto-encoder network to lossy image compression. It is only needed to encode the learned latent representation from the well-trained neural networks. However, it is not trivial to adapt auto-encoder to image compression since the original auto-encoder is not optimized for compression thereby not efficient by directly using a trained auto-encoder. In addition, there exist other major challenges: First, the low-dimension representation should be quantized before being encoded, but the quantization is not differentiable, which is required in backpropagation while training the neural networks. Second, the objective under compression scenario is different since both the distortion and the rate need to be take into consideration. Estimating the rate is challenging. Third, a practical image coding scheme needs to support variable rate, scalability, encoding/decoding speed, interoperability. In response to these challenges, a number of researchers have been actively contributing to this area. The prototype auto-encoder for image compression is in Fig. 1B, which can be regarded as a transform coding strategy. The original image ^^ is transformed with the analysis network ^^ ൌ 9 F1250625PCT ^^^^^^^, where ^^ is the latent representation which will be quantized and coded. The synthesis network will inversely transform the quantized latent representation ^ ^^ back to obtain the reconstructed image ^ ^^ ൌ ^^^^^^^^. The framework is trained with the rate-distortion loss function, i.e., ℒ ൌ ^^ ^ ^^^^, where ^^ is the distortion between ^^ and ^ ^^, ^^ is the rate calculated or estimated from the quantized representation^ ^^, and ^^ is the Lagrange multiplier. It should be noted that ^^ can be calculated in either pixel domain or perceptual domain. All existing research works follow this prototype and the difference might only be the network structure or loss function. 2.3.3 Hyper prior model In the transform coding approach to image compression, the encoder subnetwork (section 2.3.2) transforms the image vector x using a parametric analysis transform ^^^^^^,∅^^ into a latent representation ^^ , which is then quantized to form ^ ^^ . Because ^ ^^ is discrete-valued, it can be losslessly compressed using entropy coding techniques such as arithmetic coding and transmitted as a sequence of bits. As evident from the middle left and middle right image of Fig. 2, there are significant spatial dependencies among the elements of ^ ^^. Notably, their scales (middle right image) appear to be coupled spatially. An additional set of random variables ^^^ can be introduced to capture the spatial dependencies and to further reduce the redundancies. In this case the image compression network is depicted in Fig.3. In Fig 3, the left hand of the models is the encoder ^^^ and decoder ^^^ (explained in section 2.3.2). The right-hand side is the additional hyper encoder ℎ^ and hyper decoder ℎ^ networks that are used to obtain ^^^. In this architecture the encoder subjects the input image x to ^^^, yielding the responses ^^ with spatially varying standard deviations. The responses ^^ are fed into ℎ^ , summarizing the distribution of standard deviations in ^^. ^^ is then quantized (^^^), compressed, and transmitted as side information. The encoder then uses the quantized vector ^^^ to estimate ^^, the spatial distribution of standard deviations, and uses it to compress and transmit the quantized image representation^ ^^. The decoder first recovers ^^^ from the compressed signal. It then uses ℎ^ to obtain ^^, which provides it with the correct probability estimates to successfully recover ^^^ as well. It then feeds ^ ^^ into ^^^ to obtain the reconstructed image. When the hyper encoder and hyper decoder are added to the image compression network, the spatial redundancies of the quantized latent ^ ^^ are reduced. The rightmost image in Fig. 2 correspond to the quantized latent when hyper encoder/decoder are used. Compared to middle right image, the spatial redundancies are significantly reduced, as the samples of the quantized latent are less correlated. In Fig. 2: Left: an image from the Kodak dataset. Middle left: visualization of a latent representation y of that image. Middle right: standard deviations ^^ of the latent. Right: latents y 10 F1250625PCT after the hyper prior (hyper encoder and decoder) network is introduced. Fig.3 illustrates Network architecture of an autoencoder implementing the hyperprior model. The left side shows an image autoencoder network, the right side corresponds to the hyperprior subnetwork. The analysis and synthesis transforms are denoted as ^^^ and ^^^ . Q represents quantization, and AE, AD represent arithmetic encoder and arithmetic decoder, respectively. The hyperprior model consists of two subnetworks, hyper encoder (denoted with ℎ^ ) and hyper decoder (denoted with ℎ^). The hyper prior model generates a quantized hyper latent (^^^) which comprises information about the probability distribution of the samples of the quantized latent ^ ^^. ^^^ is included in the bitsteam and transmitted to the receiver (decoder) along with ^ ^^. 2.3.4 Context model Although the hyper prior model improves the modelling of the probability distribution of the quantized latent ^ ^^, additional improvement can be obtained by utilizing an autoregressive model that predicts quantized latents from their causal context (Context Model). The term auto-regressive means that the output of a process is later used as input to it. For example the context model subnetwork generates one sample of a latent, which is later used as input to obtain the next sample. Fig. 4 is a schematic diagram illustrating an example combined model configured to jointly optimize a context model along with a hyperprior and the autoencoder. The following table illustrates meaning of different symbols. Table – Illustration of symbols Component Symbol A joint arch per encoder and 11 F1250625PCT hyper decoder) and a context model subnetwork are utilized. The hyper prior and the context model are combined to learn a probabilistic model over quantized latents^ ^^, which is then used for entropy coding. As depicted in Fig. 4, the outputs of context subnetwork and hyper decoder subnetwork are combined by the subnetwork called Entropy Parameters, which generates the mean ^^ and scale (or variance) ^^ parameters for a Gaussian probability model. The gaussian probability model is then used to encode the samples of the quantized latents into bitstream with the help of the arithmetic encoder (AE) module. In the decoder the gaussian probability model is utilized to obtain the quantized latents ^ ^^ from the bitstream by arithmetic decoder (AD) module. Fig 4 illustrates the combined model jointly optimizes an autoregressive component that estimates the probability distributions of latents from their causal context (Context Model) along with a hyperprior and the underlying autoencoder. Real-valued latent representations are quantized (Q) to create quantized latents (^^^) and quantized hyper-latents (^^^ ), which are compressed into a bitstream using an arithmetic encoder (AE) and decompressed by an arithmetic decoder (AD). The highlighted region corresponds to the components that are executed by the receiver (i.e. a decoder) to recover an image from a compressed bitstream. Typically the latent samples are modeled as gaussian distribution or gaussian mixture models (not limited to). According to Fig.4, the context model and hyper prior are jointly used to estimate the probability distribution of the latent samples. Since a gaussian distribution can be defined by a mean and a variance (aka sigma or scale), the joint model is used to estimate the mean and variance (denoted as ^^ and ^^). 2.3.5 Gained variational autoencoders (G-VAE) Typically, neural network-based image/video compression methodologies need to train multiple models to adapt to different rates. Gained variational autoencoders (G-VAE) is the variational autoencoder with a pair of gain units , which is designed to achieve continuously variable rate adaptation using a single model. It comprises of a pair of gain units, which are typically inserted to the output of encoder and input of decoder. The output of the encoder is defined as the latent representation ^^ ∈ ^^^∗^∗௪, where ^^, ℎ,^^ represent the number of channels, the height and width of the latent Each channel of the latent representation is denoted as ^^^^^ ∈ ^^^∗௪, where ^^ ൌ 0, 1, … , ^^ െ 1. A pair of gain units include a gain matrix ^^ ∈ ^^^∗^ and inverse gain matrix, where ^^ is the number of gain vectors. The gain vector can be denoted as ^^^ ൌ ^^^^^^^,^^^^^^, … ,^^^^^ି^^^ , ^^^^^^ ∈ ^^ where ^^ denotes the index of the gain vectors in the gain The motivation of gain matrix is similar to the quantization table in JPEG by controlling the quantization loss based on the characteristics of different channels. To apply the gain matrix to the latent representation, each channel is multiplied with the corresponding value in a gain vector. 12 F1250625PCT ^ത^^ ൌ ^^ ^ ^^^ where ^ is channel-wise multiplication, i.e., ^ത^^^^^ ൌ ^^^^^ ൈ ^^^^^^, and ^^^^^^ is the ^^-th gain value in the gain vector ^^ . The in ^∗^ ^ verse gain matrix used at the decoder side can be denoted as ^^′ ∈ ^^ , which consists of ^^ inverse gain vectors, i.e., ^^′ ൌ ^^^^^^^, ^^^^^^, … , ^^^^^ି^^^, ^^^^^^ ∈ ^^. The inverse gain process is expressed as: ′ ^^′ ^ ൌ ^^^ ^ ^^^ ′ where ^^^ is the decoded quantized latent representation and ^^^ is the inversely gained quantized latent representation, which will be fed into the synthesis network. To achieve continuous variable rate adjustment, interpolation is used between vectors. Given two ′ ′ pairs of gain vectors ^^^௧ ,^^௧ ^ and ^^^^ ,^^^ ^, the interpolated gain vector can be obtained via the following equations. ^^௩ ൌ ^^^^^^^ ∙ ^^^௧^^ି^^ ′ ′ ′ where ^^ ∈ ^^ is an interpolation the corresponding bit rate of the generated gain vector pair. Since ^^ is a real number, an arbitrary bit rate between the given two gain vector pairs can be achieved. 2.3.6 The encoding process using joint auto-regressive hyper prior model The fig 4. corresponds to the state of the art compression method. In this section and the next, the encoding and decoding processes will be described separately. The Fig. 5 depicts the encoding process. The input image is first processed with an encoder subnetwork. The encoder transforms the input image into a transformed representation called latent, denoted by ^^. ^^ is then input to a quantizer block, denoted by Q, to obtain the quantized latent (^^^ ). ^ ^^ is then converted to a bitstream (bits1) using an arithmetic encoding module (denoted AE). The arithmetic encoding block converts each sample of the^ ^^ into a bitstream (bits1) one by one, in a sequential order. The modules hyper encoder, context, hyper decoder, and entropy parameters subnetworks are used to estimate the probability distributions of the samples of the quantized latent^ ^^. the latent ^^ is input to hyper encoder, which outputs the hyper latent (denoted by ^^). The hyper latent is then quantized (^^^) and a second bitstream (bits2) is generated using arithmetic encoding (AE) module. The factorized entropy module generates the probability distribution, that is used to encode the quantized hyper latent into bitstream. The quantized hyper latent includes information about the probability distribution of the quantized latent (^^^). 13 F1250625PCT The Entropy Parameters subnetwork generates the probability distribution estimations, that are used to encode the quantized latent ^^^. The information that is generated by the Entropy Parameters typically include a mean ^^ and scale (or variance) ^^ parameters, that are together used to obtain a gaussian probability distribution. A gaussian distribution of a random variable x is defined as ^^^^^^ ൌ ^ ఙ√ଶగ ^^ିభ ^షഋ మ మ^ ^ ^ wherein the parameter ^^ is the mean or expectation of the distribution (and also its median and mode), while the parameter ^^ is its standard deviation (or variance, or scale). In order a gaussian distribution, the mean and the variance need to be determined. In an existing design, the entropy parameters module are used to estimate the mean and the variance values. The subnetwork hyper decoder generates part of the information that is used by the entropy parameters subnetwork, the other part of the information is generated by the autoregressive module called context module. The context module generates information about the probability distribution of a sample of the quantized latent, using the samples that are already encoded by the arithmetic encoding (AE) module. The quantized latent^ ^^ is typically a matrix composed of many samples. The samples can be indicated using indices, such as ^ ^^[i,j,k] or^ ^^[i,j] depending on the dimensions of the matrix^ ^^. The samples ^ ^^[i,j] are encoded by AE one by one, typically using a raster scan order. In a raster scan order the rows of a matrix are processed from top to bottom, wherein the samples in a row are processed from left to right. In such a scenario (wherein the raster scan order is used by the AE to encode the samples into bitstream), the context module generates the information pertaining to a sample ^ ^^[i,j], using the samples encoded before, in raster scan order. The information generated by the context module and the hyper decoder are combined by the entropy parameters module to generate the probability distributions that are used to encode the quantized latent^ ^^ into bitstream (bits1). Finally the first and the second bitstream are transmitted to the decoder as result of the encoding process. It is noted that the other names can be used for the modules described above. In the above description, the all of the elements in Fig. 5 are collectively called encoder. The analysis transform that converts the input image into latent representation is also called an encoder (or auto-encoder). 2.3.7 The decoding process using joint auto-regressive hyper prior model The Fig.6 depicts the decoding process separately. In the decoding process, the decoder first receives the first bitstream (bits1) and the second bitstream (bits2) that are generated by a corresponding encoder. The bits2 is first decoded by the arithmetic decoding (AD) module by utilizing the probability distributions generated by the 14 F1250625PCT factorized entropy subnetwork. The factorized entropy module typically generates the probability distributions using a predetermined template, for example using predetermined mean and variance values in the case of gaussian distribution. The output of the arithmetic decoding process of the bits2 is ^^^, which is the quantized hyper latent. The AD process reverts to AE process that was applied in the encoder. The processes of AE and AD are lossless, meaning that the quantized hyper latent ^^^ that was generated by the encoder can be reconstructed at the decoder without any change. After obtaining of ^^^ , it is processed by the hyper decoder, whose output is fed to entropy parameters module. The three subnetworks, context, hyper decoder and entropy parameters that are employed in the decoder are identical to the ones in the encoder. Therefore the exact same probability distributions can be obtained in the decoder (as in encoder), which is essential for reconstructing the quantized latent ^^^ without any loss. As a result the identical version of the quantized latent ^^^ that was obtained in the encoder can be obtained in the decoder. After the probability distributions (e.g. the mean and variance parameters) are obtained by the entropy parameters subnetwork, the arithmetic decoding module decodes the samples of the quantized latent one by one from the bitstream bits1. From a practical standpoint, autoregressive model (the context model) is inherently serial, and therefore cannot be sped up using techniques such as parallelization. Finally the fully reconstructed quantized latent ^^^ is input to the synthesis transform (denoted as decoder in Fig.6) module to obtain the reconstructed image. In the above description, the all of the elements in Fig. 6 are collectively called decoder. The synthesis transform that converts the quantized latent into reconstructed image is also called a decoder (or auto-decoder). 2.4 Neural networks for video compression Similar to conventional video coding technologies, neural image compression serves as the foundation of intra compression in neural network-based video compression, thus development of neural network-based video compression technology comes later than neural network-based image compression but needs far more efforts to solve the challenges due to its complexity. Starting from 2017, a few researchers have been working on neural network-based video compression schemes. Compared with image compression, video compression needs efficient methods to remove inter- picture redundancy. Inter-picture prediction is then a crucial step in these works. Motion estimation and compensation is widely adopted but is not implemented by trained neural networks until recently. Studies on neural network-based video compression can be divided into two categories according to the targeted scenarios: random access and the low-latency. In random access case, it requires the decoding can be started from any point of the sequence, typically divides the entire sequence 15 F1250625PCT into multiple individual segments and each segment can be decoded independently. In low-latency case, it aims at reducing decoding time thereby usually merely temporally previous frames can be used as reference frames to decode subsequent frames. 2.5 Preliminaries Almost all the natural image/video is in digital format. A grayscale digital image can be represented by ^^ ∈ ^^^ൈ^, where ^^ is the set of values of a pixel, ^^ is the image height and ^^ is the image width. For example, ^^ ൌ ^0, 1, 2, … ,255^ is a common setting and in this case |^^| ൌ 256 ൌ 2଼, thus the pixel can be represented by an 8-bit integer. An uncompressed grayscale digital image has 8 bits-per-pixel (bpp), while compressed bits are definitely less. A color image is typically represented in multiple channels to record the color information. For example, in the RGB color space an image can be denoted by ^^ ∈ ^^^ൈ^ൈଷ with three separate channels storing Red, Green and Blue information. Similar to the 8-bit grayscale image, an uncompressed 8-bit RGB image has 24 bpp. Digital images/videos can be represented in different color spaces. The neural network-based video compression schemes are mostly developed in RGB color space while the traditional codecs typically use YUV color space to represent the video sequences. In YUV color space, an image is decomposed into three channels, namely Y, Cb and Cr, where Y is the luminance component and Cb/Cr are the chroma components. The benefits come from that Cb and Cr are typically down sampled to achieve pre-compression since human vision system is less sensitive to chroma components. A color video sequence is composed of multiple color images, called frames, to record scenes at different timestamps. For example, in the RGB color space, a color video can be denoted by ^^ ൌ ^^^ , ^^ , … , ^^ , … ^ ^ൈ^ ^ ^ ௧ , ^^்ି^ where ^^ is the number of frames in this video sequence, ^^ ∈ ^^ . If ^^ ൌ 1080, ^^ ൌ 1920, |^^| ൌ 2଼, and the video has 50 frames-per-second (fps), then the data rate of this uncompressed video is 1920 ൈ 1080 ൈ 8 ൈ 3 ൈ 50 ൌ 2,488,320,000 bits-per-second (bps), about 2.32 Gbps, which needs a lot storage thereby definitely needs to be compressed before transmission over the internet. Usually the lossless methods can achieve compression ratio of about 1.5 to 3 for natural images, which is clearly below requirement. Therefore, lossy compression is developed to achieve further compression ratio, but at the cost of incurred distortion. The distortion can be measured by calculating the average squared difference between the original image and the reconstructed image, i.e., mean-squared-error (MSE). For a grayscale image, MSE can be calculated with the following equation. ^^^^^^ ൌ ‖௫ି௫^‖మ (4) Accordingly, the quality of the 16 F1250625PCT measured by peak signal-to-noise ratio (PSNR): ^^^^^^^^ ൌ 10 ൈ ^^^^^^ ^^^௫^^^^^మ ^^ ெௌா (5) where ^^^^^^^^^^ is the quality evaluation metrics SSIM). To compare different lossless compression schemes, it is sufficient to compare either the compression ratio given the resulting rate or vice versa. However, to compare different lossy compression methods, it has to take into account both the rate and reconstructed quality. For example, to calculate the relative rates at several different quality levels, and then to average the rates, is a commonly adopted method; the average relative rate is known as Bjontegaard’s delta- rate (BD-rate). There are other important aspects to evaluate image/video coding schemes, including encoding/decoding complexity, scalability, robustness, and so on. 2.6 Separate processing of luma and chroma components of an image Fig. 7 illustrates the decoding process according to some example embodiments of the present disclosure. According to one implementation, the luma and chroma components of an image can be decoded using separate subnetworks. In Fig.7, the luma component of the image is processed by the subnetwoks “Synthesis”, “Prediction fusion”, “Mask Conv”, “Hyper Decoder”, “Hyper scale decoder” etc. Whereas the chroma components are processed by the subnetworks: “Synthesis UV”, “Prediction fusion UV”, “Mask Conv UV”, “Hyper Decoder UV”, “Hyper scale decoder UV” etc. A benefit of the above separate processing is that the computational complexity of the processing of an image is reduced by application of separate processing. Typically in neural network based image and video decoding, the computational complexity is proportional to the square of the number of feature maps. If the number of total feature maps is equal to 192 for example, computational complexity will be proportional to 192x192. On the other hand if the feature maps are divided into 128 for luma and 64 for chroma (in the case of separate processing), the computational complexity is proportional to 128x128 + 64x64, which corresponds to a reduction in complexity by 45%. Typically the separate processing of luma and chroma components of an image does not result in a prohibitive reduction in performance, as the correlation between the luma and chroma components are typically very small. The processing (Decoding process) in Fig.7 can be explained below: 1. Firstly, the factorized entropy model is used to decode the quantized latents for luma and chroma, i.e., ^^^ and ^^^^^^^ in Fig.7. 2. The probability parameters (e.g. variance) generated by the second network are used to generate a quantized residual latent by performing the arithmetic decoding process. 17 F1250625PCT 3. The quantized residual latent is inversely gained with the inverse gain unit (iGain) as shown in orange color in Fig. 7. The outputs of the inverse gain units are denoted as ^^^ and ^^^^^^^ for luma and chroma components, respectively. 4. For the luma component, the following steps are performed in a loop until all elements of ^ ^^ are obtained: a. A first subnetwork is used to estimate a mean value parameter of a quantized latent (^^^), using the already obtained samples of ^ ^^. b. The quantized residual latent ^^^ and the mean value are used to obtain the next element of ^ ^^. 5. After all of the samples of ^ ^^ are obtained, a synthesis transform can be applied to obtain the reconstructed image. 6. For chroma component, step 4 and 5 are the same but with a separate set of networks. 7. The decoded luma component is used as additional information to obtain the chroma component. Specifically, the Inter Channel Correlation Information filter sub-network (ICCI) is used for chroma component restoration. The luma is fed into the ICCI sub- network as additional information to assist the chroma component decoding. 8. Adaptive color transform (ACT) is performed after the luma and chroma components are reconstructed. The module named ICCI is a neural-network based postprocessing module. The example embodiments of the present disclosure are not limited to the UCCI subnetwork, any other neural network based postprocessing module might also be used. An exemplary implementation of some example embodiments of the present disclosure is depicted in Fig. 7 (the decoding process). The framework comprises two branches for luma and chroma components respectively. In each of the branch, the first subnetwork comprises the context, prediction and optionally the hyper decoder modules. The second network comprises the hyper scale decoder module. The quantized hyper latent are ^^^ and ^^^^^^^. The arithmetic decoding process generates the quantized residual latents, which are further fed into the iGain units to obtain the gained quantized residual latents ^^^ and ^^^^^^^. After the residual latent is obtained, a recursive prediction operation is performed to obtain the latent ^ ^^ and ^ ^^^^^^. The following steps describe how to obtain the samples of latent ^ ^^^: , ^^, ^^^, and the chroma component is processed in the same way but with different networks. 1. An autoregressive context module is used to generate first input of a prediction module using the samples ^ ^^^: , ^^,^^^ where the (m, n) pair are the indices of the samples of the latent that are already obtained. 2. Optionally the second input of the prediction module is obtained by using a hyper decoder 18 F1250625PCT and a quantized hyper latent^ ^^^^. 3. Using the first input and the second input, the prediction module generates the mean value ^^^^^^^^^: , ^^, ^^^. 4. The mean value ^^^^^^^^^: , ^^, ^^^ and the quantized residual latent ^^^^: , ^^, ^^^are added together to obtain the latent ^ ^^^: , ^^, ^^^. 5. The steps 1-4 are repeated for the next sample. Whether to and/or how to apply at least one method disclosed in the document may be signaled from the encoder to the decoder, e.g. in the bitstream. Alternatively, whether to and/or how to apply at least one method disclosed in the document may be determined by the decoder based on coding information, such as dimensions, color format, etc. Alternative or additionally, the modules named MS1, MS2 or MS3+O (in Fig. 7), might be included in the processing flow. The said modules might perform an operation to their input by multiplying the input with a scalar or adding an adding an additive component to the input to obtain the output. The scalar or the additive component that are used by the said modules might be indicated in a bitstream. The module named RD or the module named AD in Fig.7 might be an entropy decoding module. It might be a range decoder or an arithmetic decoder or the like. The example embodiments of the present disclosure described herein are not limited to the specific combination of the units exemplified in Fig. 7. Some of the modules might be missing and some of the modules might be displaced in processing order. Also additional modules might be included. For example: 1. The ICCI module might be removed. In that case the output of the Synthesis module and the Synthesis UV module might be combined by means of another module, that might be based on neural networks. 2. One or more of the modules named MS1, MS2 or MS3+O might be removed. The core of the proposed solution is not affected by the removing of one or more of the said scaling and adding modules. In Fig. 7, other operations that are performed during the processing of the luma and chroma components are also indicated using the star symbol. These processes are denoted as MS1, MS2, MS3+O. These processing might be, but not limited to, adaptive quantization, latent sample scaling, and latent sample offsetting operations. For example, in an adaptive quantization process might correspond to scaling of a sample with multiplier before the prediction process, wherein the multiplier is predefined or whose value is indicated in the bitstream. The latent scaling process might correspond to the process where a sample is scaled with a multiplier after the prediction process, wherein the value of the multiplier is either predefined or indicated in the bitstream. The 19 F1250625PCT offsetting operation might correspond to adding an additive element to the sample, again wherein the value of the additive element might be indicated in the bitstream or inferred or predetermined. Another operation might be tiling operation, wherein samples are first tiled (grouped) into overlapping or non-overlapping regions, wherein each region is processed independently. For example the samples corresponding to the luma component might be divided into tiles with a tile height of 20 samples, whereas the chroma components might be divided into tiles with a tile height of 10 samples for processing. Another operation might be application of wavefront parallel processing. In wavefront parallel processing, a number of samples might be processed in parallel, and the amount of samples that can be processed in parallel might be indicated by a control parameter. The said control parameter might be indicated in the bitstream, be inferred, or can be predetermined. In the case of separate luma and chroma processing, the number of samples that can be processed in parallel might be different, hence different indicators can be signalled in the bitstream to control the operation of luma and chrome processing separately. 2.7 Colors separation and conditional coding In one example the primary and secondary color components of an image are coded separately, using networks with similar architecture, but different number of channels as shown in 8. All boxes with same names are sub-networks with the similar architecture, only input-output tensor size and number of channels are different. Number of channels for primary component is ^^^ ൌ 128, for secondary components is ^^^ ൌ 64. The vertical arrows (with arrowhead pointing downwards) indicate data flow related to secondary color components coding. Vertical arrows show data exchange between primary and secondary components pipelines. The input signal to be encoded is notated as ^^, latent space tensor in bottleneck of variational auto- encoder is ^^. Subscript “Y” indicates primary component, subscript “UV” is used for concatenated secondary components, there are chroma components. Fig.8 illustrates learning-based image codec architecture. First the input image that has RGB color format is converted to primary (Y) and secondary components(UV). The primary component ^^^ is coded independently from secondary components ^^^^ and the coded picture size is equal to input/decoded picture size. The secondary components are coded conditionally, using ^^^ as auxiliary information from primary component for encoding ^^^^ and using ^^^^ as a latent tensor with auxiliary information from primary component for decoding ^^^^^ reconstruction. The codec structure for primary component and secondary components are almost identical except the number of channels, size of the channels and the several entropy models for transforming latent tensor to bitstream, therefore primary and secondary latent tensor will generate two different bitstream based on two different entropy models. 20 F1250625PCT Prior to the encoding ^^^, ^^^^ goes through a module which adjusts the sample location by down- sampling (marked as “s^” on Fig.1), this essentially means that coded picture size for secondary component is different from the coded picture size for primary component. The scaling factor s is variable, but the default scaling factor is ^^ ൌ 2. The size of auxiliary input tensor in conditional coding is adjusted in order the encoder receives primary and secondary components tensor with the same picture size. After reconstruction, the secondary component is rescaled to the original picture size with a neural-network based upsampling filter module (“NN-color filter s^” on Fig. 1), which outputs secondary components up-sampled with factor ^^. The example in Fig. 8 exemplifies an image coding system, where the input image is first transformed into primary (Y) and secondary components (UV). The outputs ^^^^ , ^^^^^ are the reconstructed outputs corresponding to the primary and secondary components. At the and of the processing, ^^^^, ^^^^^ are converted back to RGB color format. Typically the ^^^^ is downsampled (resized) before processing with the encoding and decoding modules (neural networks). For example the size of the ^^^^ might be reduced by a factor of 50% in each of the vertical and horizontal dimensions. Therefore the processing of the secondary component includes approximately 50% x 50% = 25% less samples, therefore it is computationally less complex. 2.8 Cropping operation in neural network based coding Fig. 9 illustrates synthesis transform example for learning based image coding. The example synthesis transform above includes a sequence of 4 convolutions with up-sampling with stride of 2. The synthesis transform sub-Net is depicted on Fig.9. The size of the tensor in different parts of synthesis transform before cropping layer is the diagram on Fig.9. The cropping layer changes tensor size ℎௗ ൈ ^^ௗ to ℎௗି^ ൈ ^^ௗି^ , where ℎௗ ൌ 2 ∙ ^^^^^^^^^^^/ 2ௗ^; ^^ௗ ൌ 2 ∙ ^^^^^^^^^^^/2ௗ^ ; here ^^ is the depth of proceeding convolution in the codec architecture. For primary component Synthesis Transform receives input tensor with sizeℎ ൈ ^^; ℎ ൌ ^^^^^^^^^^^⁄ 16 ^; ^^ ൌ ^^^^^^^^^^^⁄ 16 ^. The output of Synthesis Transform for primary component is 1 ൈ ℎ^ ൈ ^^^ , where ℎ^ ൌ ^^;ℎ^ ൌ ^^. For secondary component Synthesis Transform receives input tensor with size ℎ^^ ൈ ^^^^ ; ℎ^^ ൌ ^^^^^^^^^^^^^^^^^^^^⁄ ^^ ^⁄ 16 ^; ^^^^ ൌ ^^^^^^^^^^^^^^^^^^^^⁄ ^^ ^⁄ 16 ^. The output of the Synthesis Transform for primary component is 2 ൈ ℎ^^^ ൈ ^^^^^ , where ℎ^^^ ൌ ^^^^^^^^^^^⁄ ^^ ^;ℎ^^^ ൌ ^^^^^^^^^^^⁄ ^^ ^. For secondary components input sizes are ℎ^ ൌ ^^^^^^^^^^^/^^^; ^^^ ൌ ^^^^^^^^^^^/^^^, where ^^ is the scale factor. The scale factor might be 2 for example, wherein the secondary component is downsampled by a factor of 2. Based on the above explanation, the operation of the cropping layers depend on the output size H,W and the depth of the cropping layer. The depth of the left-most cropping layer in Fig. 9 is equal to 0. The output of this cropping layer must be equal to H, W (the output size), if the size 21 F1250625PCT of the input of this cropping layer is greater than H or W in horizontal or vertical dimension respectively, cropping needs to be performed in that dimension. The second cropping layer counting from left to right has a depth of 1. The output of the second cropping layer must be equal to ℎ^ ൌ 2 ∙ ^^^^^^^^^^^/2^^; ^^^ ൌ 2 ∙ ^^^^^^^^^^^/2^^, which means if the input of this second cropping layer is greater than h1, w1 in any dimension, than cropping is applied in that dimension. In the operation of cropping layers are controlled by the output size H,W. In one example if H and W are both equal to 16, then the cropping layers do not perform any cropping. On the other hand if H and W are both equal to 17, then all 4 cropping layers are going to perform cropping. 2.9 Bitwise shifting The bitwise shift operator can be represented using the function ^^^^^^^^ℎ^^^^^^^^^,^^^, where n is an integer number. If n is greater than 0, it corresponds to right-shift operator (>>), which moves the bits of of the input to the right, and the left-shift operator (<<), which moves the bits to the left. In another words the ^^^^^^^^ℎ^^^^^^^^^,^^^ operation corresponds to: ^^^^^^^^ℎ^^^^^^^^^,^^^ ൌ ^^ ∗ 2^, or ^^^^^^^^ℎ^^^^^^^^^,^^^ ൌ ^^^^^^^^^^^^^ ∗ 2^^, or ^^^^^^^^ℎ^^^^^^^^^,^^^ ൌ ^^//2^. The output of the bitshift operation is an integer value. In some implementations, the floor() function might be added to the definition. Floor( x ) is equal to the largest integer less than or equal to x. The “//” operator or the integer division operator: It is an operation that comprises division and truncation of the result toward zero. For example, 7 / 4 and −7 / −4 are truncated to 1 and −7 / 4 and 7 / −4 are truncated to −1. ^^^^^^ℎ^^^^ℎ^^^^^^^^^,^^^ ൌ ^^ ≫ ^^ or ^^^^^^^^^^ℎ^^^^^^^^^,^^^ ൌ ^^ ≪ ^^ Equation 3: alternative implementation of the bitshift operator as rightshift or leftshift. x >> y Arithmetic right shift of a two's complement integer representation of x by y binary digits. This function is defined only for non-negative integer values of y. Bits shifted into the most significant bits (MSBs) as a result of the right shift have a value equal to the MSB of x prior to the shift operation. x << y Arithmetic left shift of a two's complement integer representation of x by y binary digits. This function is defined only for non-negative integer values of y. Bits shifted into the least significant bits (LSBs) as a result of the left shift have a value equal to 0. 2.10 Convolution operation The convolution is a fairly simple operation at heart: you start with a kernel, which is simply a small matrix of weights. This kernel “slides” over the input data, performing an elementwise 22 F1250625PCT multiplication with the part of the input it is currently on, and then summing up the results into a single output pixel. In some cases the convolution operation might comprise a “bias”, which is added to the output of the elementwise multiplication operation. The convolution operation might be described by the following mathematical formula. An output out1 can be obtained as: ெ ே ^ ^^^^^^1^^^,^^^ ൌ ^^^^^^^^1^^^^ ൌ ^^^^^1^^^^, ^^^ ൈ ^^^^^^ ^ ^^,^^ ^ ^^^ ^ ^^1 wherein w1 and ^^^ is the kth input, and direction. The convolution layer might consist of convolution operations wherein more than one output might be generated. Other equivalent depictions of the convolution operation might be found below: ெ ே ^ ^^^^^^1^^^,^^^ ൌ ^^^^^^^^1^^^^ ൌ ^ ^^^^1^^^, ^^, ^^^ ൈ ^^^^^, ^^ ^ ^^, ^^ ^ ^^^ ^ ^^1 In the number, out[1,x,y] one a number, ^^^1, ^^,^^^ is one input and ^^^2, ^^, ^^^ is a second input. The w1, or w describe weights of the convolution operation. 2.11 Leaky_Relu activation function The leaky relu activation function is depicted in Fig.10. According to the function, if the input is a positive value, the output is equal to the input. If the input (y) is a negative value, the output is equal to a*y. The a is typically (not limited to) a value that is smaller than 1 and greater than 0. Since the multiplier a is smaller than 1, it can be implemented either as a multiplication with a non-integer number, or with a division operation. The multiplier a might be called the negative slope of the leaky relu function. 2.12 Relu activation function The relu activation function is depicted in Fig.11. According to the function, if the input is a positive value, the output is equal to the input. If the input (y) is a negative value, the output is equal to 0. 2.13 The JPEG AI image coding standard The existing design utilizes some NN-based image coding methods described mentioned above. Some of the features are described or summarized below. The numbers in parentheses are section numbers for reference. 23 F1250625PCT Depending on input picture height ^^ and width ^^ and scaling factors for primary (^^^ ) and secondary (^^^^) components sizes of tensors are shown in Table 1. Table 1 Tensor size parameters for primary and secondary components encoding. Primary “Y” component Secondary “UV” component ൯ ൯ D secondary (^^^^) components sizes of tensors are shown in Table 2. Table 2 Tensor size parameters for primary and secondary components decoding. Primary “Y” component Secondary “UV” Depending on output picture and ratio between sizes of primary and secondary components in coded picture format up-sampling in vertical (with scaling factor ^^௩^^) or horizontal (with scaling factor ^^^^^) direction for secondary component is performed. Supported combintations of output picture format and corresponding sizes of primary (^^^ ൈ^^^) and secondary (^^^^ ൈ^^^^), also 24 F1250625PCT component scaling factors are listed in Table 3. Table 3 Supported combinations of output picture format and scaling factors output picture color sampling color sampling mode mode coded ^^௩^^ ^^^^^ ^^௩^^ ^^^^^ . ( . . ) pa ng ayer padding layer is denoted as Padd (h,w). Padding layer receives tensor input ^C,hin,win^ and outputs a tensor output of size ^C,h,w^ (hin^ h, win^ w). By default padding is performed by replication: For c=0, …, C, i=0,…,h, j=0,…,w, 1,j^w)?j:w-1^. Different model of padding can be specified (for example, padding by zeros). . (3.6.20) cropping layer cropping layer is denoted as Crop (h,w). Cropping layer receives tensor input of size ^C,hin,win^ and outputs a tensor output of size ^C,h,w ^ (h^ hin, w^ win). Cropping is performed by discarding redundant elements: For c=0, …, C, i=0,…,h, j=0,…,w, . (3.6.23) Down-shuffle operation Down-shuffle operation is specific type of unshuffle operation with scaling factor 2. 25 F1250625PCT The input of this process is: ^ ^^^^^, 2ℎ, 2^^^ - 3D tensor. The output of this process is: ^ ^^^ ^4^^, ℎ,^^^ re-shuffled 3D tensor with same elements. Since down-shuffle operation changes spatial size of tensor similar way as down-sampling convolution with stride 2, this process is preceeded by padding layer (2h,2w). Zero padding is performed. Down-shuffle process is as follows: For c=0,…C-1 , i=0,…, ℎ െ1, j=0,…, ^^ െ1, ^ ^^^ ^4^^ ^ 4^^ , ^^, ^^^ ൌ ^^^^^ ^ ^^, 2^^ ,2^^ ^, ^ ^^^ ^4^^ ^ 4^^ ^ 1, ^^, ^^^ ൌ ^^^^^ ^ ^^, 2^^ ^ 1,2^^ ^ 1^, ^ ^^^ ^ ^ ^^^ ^^^^^ ^ ^ ^, ^. This process is illustrated in the Fig.12. 4. (3.6.24) Up-shuffle operation This process is inverse to down-shuffle operation, specific type of shuffle operation with scaling factor 2. The input of this process is: ^ ^^^ ^4^^, ℎ,^^^ 3D tensor. The output of this process is: ^ re-shuffled 3D tensor ^^^^^, 2ℎ, 2^^^ with same elements. For c=0,…C-1 , i=0,…, ℎ െ1, j=0,…, ^^ െ1, ^ ^^^^^ ^ ^^, 2^^ ,2^^ ^ ൌ ^^^ ^4^^ ^ 4^^ , ^^, ^^^, ^ ^^^^^ ^ ^^, 2^^ ^ 1,2^^ ^ 1^ ൌ ^^^ ^4^^ ^ 4^^ ^ 1, ^^, ^^^, ^^^, ^ ^^^^^ ^ ^^, 2^^ ^ 1,2^^ ^ ൌ ^^^ ^4^^ ^ 4^^ ^ 3, ^^, ^^^. This process is illustrated in the Fig.13. Since up-shuffle operation changes spatial size of tensor similar way as inverse convolution with stride 2, this process is followed by cropping layer (2h,2w). 5. (5.4) Latent space tiles Latent space tiles decoding can be enabled for each component independently by flags signalled in picture header (section 9.3). If tile_enable_Luma or tile_enable_Choma is true then corresponsing component is decoded unsing latent tiling process illustrated on Fig.14. In this section Hin = H, Win = W for primary component and Hin = H UV =H/cver , Win = WUV 26 F1250625PCT =W/chor for secondary component. The number and location of tiles are determined by values tile size Stile (equal to tile_size_Luma for primary component and tile_size_Chroma for secondary component) and tile overlap mtile (tile_ overlap _Luma for primary component tile_ overlap_Chroma for secondary component ) signalled in picture header (section 9.3). Further, the ratio ^^ ൌ ^^^^⁄ ℎସ ൌ ^^^^⁄ ^^ସ between signal domain size and latent tensor size is known, which is ^^ ൌ 2ସ for primary and ^^ ൌ 2ଷ for secondary component. The number of tiles vertically is ^^ ൌ ^^^^^^^^ ^^^^^^ െ ^^௧^^^^⁄ ^ ^^௧^^^ െ ^^௧^^^^ ^ and horizontally ^^ ൌ ^^^^^^^^ ^^^^^^ െ ^^௧^^^^⁄ ^^^௧^^^ where ^^^^,^^^^ are height and width of output color plane. ^^^^,^^^^ are hight and width plane. For ^^ ൌ 0, … ,^^ െ 1 and ^^ ൌ 0, … , ^^ െ 1, - Define tiles coordinates and sizes o ^^ ^^^ ^ ൌ ^^ ∙ ^^^௧^^^ െ ^^௧^^^^ – vertical dimension tile start in signal domain; o ^^^ ^^^ – vertical dimension tile start in latent space; o ^ ^^௧^^^^ ^ ^^^^^ ? ^^௧^^^:^^^^ െ ^^ ^^^ ^ – vertical dimension tile size in o ℎସ ൌ ^^^^^^^^^^^^ ⁄ ^^ ^– vertical dimension tile size in latent space; o ൌ ^^ ∙ ^^^௧^^^ െ ^^௧^^^^ – horizontal dimension tile start in signal domain; o ^^ ^^^ ସ ൌ ^^ ^^^ ^ ⁄ ^^ – horizontal dimension tile start in latent space; o ^^ ^^^ ^^^ ^ ^ ^ ൌ ൬^^^^ ^ ^^௧^^^^ ^ ^^^^^ ? ^^௧^^^: ^^^^ െ ^^ ^ ^ – horizontal dimension tile size in signal o ^^ ^^^ ସ ൌ ^^^^^^^^^^^ ^^^ ^ ⁄ ^^ ) – horizontal dimension tile size in latent space. - tile latent space tensor o For ^^ ൌ 0, … ,^^ െ 1 , ^^ ൌ 0, … ,ℎ ^^^ ^^^ ସ െ 1 and ^^ ൌ 0, … ,^^ସ െ 1 1 - o Synthesis transform described in section 8.3 with ^^^ ^^^ ^^^ ^,^, ^^^^,^ of sizes ℎସ and ^^ସ as inputs and ^^^ ^^^ ^^^ ^,^ of size ^^^ and ^^^ as output. 27 F1250625PCT - Merging reconstructed parts of tensor into one o ov୪ = ^^^ ^^^ ^ ൌൌ 0 ^? 0:^^௧^^^ / 2 – overlap used on left boundary of the tile; ^^^ of o ୠ = ^ ^ – of the tile; ^ ^ ^ ^ o ^^ ^ ^ ൌ ^^ ^ ^ െ ov^ െ ovୠ vertical dimension tile size in signal domain without tile o ^ ൌ ^^ ^^^ ^ െ ov୪ െ ov୰ horizontal dimension tile size in signal domain without o For ^^ ൌ 0, … ,^^ െ 1 , ^^ ൌ 0, … ,^^ ^^^ െ 1 a ^^^ ^^ ^ nd ^^ ൌ 0, … , ^^^ െ 1, ^^^ ^^^ ^^^. Tile size . (9.2) Code stream layout Code stream structure is depicted in Fig. 15. The code stream is composed of six parts with byte boundary, which are: 1. SOC - Start Of Codestream marker; 2. PIH (Picture Header marker) followed by picture header (section 9.3); 3. TOH (Tools Header marker) followed by tools information (section 9.3.1.1); 4. SOQ (start of Quality map marker) followed by code stream; 5. SOZ (Start of Z-stream marker) followed codestream of hyper tensor z, including ^̂^^ ("^^^^^^^^^^^^ ^^^" in Fig.16) and ^̂^^^("^^^^^^^^^^^^ ^^^^" in Fig.16); 6. SORp (Start of Residual stream for primary component marker) followed by codestream of primary component residual, which includes ^̂^^ ("^^^^^^^^^^^^ ^^^" in Fig.16); 7. SORs (Start of Residual stream for secondary component marker) followed by codestream of secondary component residual, which includes ^̂^^^ ("^^^^^^^^^^^^ ^^^^" in Fig.16); 8. EOC - End Of Codestream marker. The overall syntax structure of an image is: picture() { Descriptor 28 F1250625PCT picture_header() tools_header() E ows: Code Symbol Description Mandatory/Optional assi nment . (6.3.1) Syntax table 29 F1250625PCT picture_header( ) { Descriptor PIH u(16) (...) o e ea er model_header(comp) { Descriptor skip_mode_header(comp) { Descriptor 30 F1250625PCT cube_flag_header (comp) { Descriptor NumCubeFlag =4*( (img_height + 15) >> 4) * ( (img_width + 15) >> 4) e_eae co p { esc por color_transform_header() { Descriptor 31 F1250625PCT . (6.3.2) Picture header semantics Following service information is signalled: picture_header_size is the number of bytes in the picture header excluding the first two-byte marker; img_width plus 64 specifies width of an input picture (from 64 to 65599); img_height plus 64 specifies height of the input picture (from 64 to 65599); bit_depth is a bit-depth the output picture (“0” corresponds to 8 and “1” corresponds to 10); s_ver_minus1 is one bit value which defines sver = 1 + s_ver_minus1; sver is a ratio between primary and secondary components height in output picture. Allowed values for sver are specified in Table 2. Usage of sver is descried in sections 4.4, 7.6, 7.7, 7.8. s_hor_minus1 is a one bit value which defines shor =1+ s_hor_minus1, shor is a ratio between primary and secondary components width in output picture. Allowed values for sver are specified in Table 2. Usage of sver is descried in sections 4.4, 7.6, 7.7, 7.8. c_ver_minus1 is one bit value which defines cver = 1 + c_ver_minus1; cver is a ratio between primary and secondary components height in coded picture. Allowed values for cver are specified in Table 2: 1^ sver^cver ^2. Usage of cver is descried in the sections 4.4 , 7.6, 8.3. c_hor_minus1 is a one bit value which defines chor =1+ c_hor_minus1, chor is a ratio between primary and secondary components width in coded picture. Allowed values for chor are specified in Table 2: 1^ shor^chor ^2. . Usage of chor is descried in the sections sections 4.4 , 7.6, 8.3. independent_beta_uv is a flag (false/true), independent_beta_uv equal to true indicates that the beta displacement parameter for primary and secondary components are different. If independent_beta_uv equal to false then beta displacement parameter for primary and secondary components are the same. beta_displacement_log_plus_2048[comp] minus 2048 is a – parameter indicating a displacement between the rate control parameter beta selected by encoder for comp component and the reference rate control parameter beta associated with the index of the used model (model_id). The displacement is in logarithmic scale. betaDisplacementLog[comp] = beta_displacement_log_plus_2048[comp] – 211 NOTE – the reference rate control parameter beta (β), mentioned here, is the parameter used in the model training to control the ratio between the bitrate and distortion. “The model” here means the model associated with the index of the used model (model_id). When syntax element beta_displacement_log_plus_2048[1] is not present (independent_beta_uv is equal to 0), betaDisplacementLog[1] = betaDisplacementLog[0]. 32 F1250625PCT model_id is an identificator of pre-stored checkpoint with model’s weights, model_id = 0,1,2 or 3. decoderID is an identificator for synthesis transfrom net (8.3). tile_enable[comp] are enable flags for tiling of primary (comp==0) and secondary (comp==1) components. tile_size[comp] are size of tiles for primary (comp==0) and secondary (comp==1) components. tile_overlap[comp] are sizes of tiles overlapping areas for primary (comp==0) and secondary (comp==1) components. cube_group_flag[comp] is 1-bit unsigned integer. cube_group_ flag=0 indicates no cube_flag is signalled in one group, and cube flags of one group are set to be 1. cube_group_ flag=1 indicates cube flags of one group are signaled. cube_flag[comp] is 1D array of size ((h4+7)^^3)^((w4+7)^^3), which contains cube flags for primary component (comp =0). Value 1 indicates Skip Mode is applied to one cube of residual tensor ^̂^^^^^. Value 0 indicates Skip Mode is disable for one cube of residual tensor ^̂^^^^^. If the comp value is 0, then the flag comtrols to the primary component residual signaling, otherwise the secondary component residual signaling h4, w4 are defined in Table 1. color_transform_idx index indicating transformation between internal and output color format, color_transform_idx equal to 0 indicates no color transformation, color_transform_idx equal to 1 indicates color transformation as in ITU-R BT.709, color_transform_idx equal to 2 indicates color transformation with user defined parameters. color_transform_matrix[i][j] is a matrix of color convertion, signalled only for color_transform_idx equal to 2. color_transform_offset[i] is an offset for color convertion, signalled only for color_transform_idx equal to 2. 10. (6.4.1) Tools header tools_header( ) { Descriptor comp_tools_header(comp) { Descriptor 33 F1250625PCT lsbs_header(comp) rvs_header(comp) _ _ r EFElinearfilterparameters () _ or rvsenablefla[com] u(1) lsbs_header() { Descriptor gain_parameters(comp) { Descriptor 34 F1250625PCT } } or iccienableflag u(1) _ _ or iccitileenable u(1) ICCI_comp_model_header(comp_idx) { Descriptor 35 F1250625PCT icci_model_idx[comp_idx][tile_idx] u(4) else _ or LEFenabledflag uf(1) _ _ _ or EFElinearfilterenabledfla u(1) EFE_non_linear_filter_parameters( ) { Descripto 36 F1250625PCT len_mask_x u(10) } o 37 F1250625PCT for (i = 0, i < candNum, i++){ u(16) A1 u(16) r bestcandidx[0] u(4) decode_one_filter( numFilter, numPlane, fl, maxSymbol, minSymbol) { Descriptor PCT for (fl_hor = 0, fl_hor < fl - 1, fl_hor++) for (fl_ver = 0, fl_ver < fl - 1, fl_ver++) b b e va ues o [ ], [ ], es _can _ x[ ] an es _can _ x[ ] are res rc e o e n the first invocation of decodeFilters( ). 11. (6.4.2) Tools information semantics tools_header_size is the number of bytes in the tools header excluding the first two-byte marker; rvs_enable_flag[comp] – is a flag used in the RVS for comp component.0 indicates RVS disabled, 1 indicates RVS turn on. rvs_num_idx[comp] is an identificator for operation point, 0 means 1 threshold for RVS, 1 means 3 thresholds for RVS. lsbs_enable_flag[comp] – is a flag used in the LSBS mode for comp component.0 indicates LSBS disabled, 1 indicates LSBS turn on. grfs_enable_flag[comp] is an enable flag for channel-wise gain unit refinement scale tool, if grfs_enable_flag[comp] is equal to one then residual tensor elements in channel indicated by grfs_channel_flag[comp] for comp component are scalled. grfs_channel_flag[comp] is an array with 1-bit flags. Its size depends on value of the comp. For comp equals to 0 the size is Cp and for comp equals to 1 the size is Cs. Channels with grfs_channel_flag[comp] equal to 0 the absolute value of comp component residual tensor is 39 F1250625PCT reduced, for scales with grfs_channel_flag[comp] equal to 1 the absolute value of comp component residual tensor is amplified. Here comp==0 indicates primary component and comp==1 indicates secondary component, number of channels for primary Cp and secondary Cs component latent tensor are defined in Table 1. gain_3D_enable_flag[comp] is an enable flag for local control, if gain_3D_enable_flag[comp] is equal to 1 then residual tensor elements are scalled according to rules specified in section 12.2. quality_map_entropy_index[comp] is an indicator which defines sigma index for comp component quality map information decoding in me-tANS. Values of quality_map_entropy_index[comp] are in the range [0, 7]. q_stream () is process to decode quality map information (described in section 9.5.2). icci_enable is an enable flag for ICCI. icci_tile_enable is an enable flag for tiling of ICCI. icci_tile_size is a size of tiles for ICCI. icci_tile_overlap is a size of tiles overlapping areas for primary and secondary components. icci_use_single_model[comp_idx] is a flag that only one particular module of ICCI is used for each color component. icci_model_idx[comp_idx][ tile_idx] is an array with model’s index for each tile of each color component. icci_use_default_model[comp_idx] is a flag, which indicates whether default model index will be use for each color component. icci_default_model_idx[comp_idx] is a default model index for each color components. icci_use_default_idx[comp_idx][tile_idx] is a flag which indicates whether the default model index will be used or not. EFE_linear_filter_enabled_flag – flag indicating if EFE luma-aided chroma linear filter process is enabled. EFE_nonlinear_filter_enabled_flag – flag indicating if EFE nonlinear filtering processed is enabled. best_cand_idx[0] – the 4 bit non-negative integer value specifying the candidate index corresponding to the u-component (first one of the secondary components), indicating the number of tiles and the tile coordinates. It is used as input to cand [X][Y] table in section 14.2.2. best_cand_idx[1] – the 4 bit non-negative integer value specifying the candidate index corresponding to the v-component (second one of the secondary components), indicating the number of tiles and the tile coordinates. It is used as input to cand [X][Y] table in section 14.2.2. fl[0] – the 9-valued non-negative integer value specifying the kernel size. The value of fl[0] is 40 F1250625PCT restricted to be smaller than 4 and greater than 0. fl[1] – the 9-valued non-negative integer value specifying the kernel size. The value of fl[1] is restricted to be smaller than 4 and greater than 0. W1A – the 4-dimensinal tensor specifying the multiplier coefficients, e.g. weights, used in section 14.1.1. W1B – the 4-dimensinal tensor specifying the multiplier coefficients, e.g. weights, used in section 14.1.1. W4A – the 4-dimensinal tensor specifying the multiplier coefficients, e.g. weights, used in section 14.1.1. W4B – the 4-dimensinal tensor specifying the multiplier coefficients, e.g. weights, used in section 14.1.1. bS – the 10 bit non-negative integer value specifying the block size of the EFE output adjustment subprocess. minSymbol – the 16-bit non-negative integer value specifying the the minimum value that is used in deinteger() function. maxSymbol – the 16-bit non-negative integer value specifying the maximum coefficient value. mask1_enabled_flag – the 1-bit non-negative integer value specifying a if the values of len_mask_1_x and len_mask_1_y are zero or greater than zero. mask2_enabled_flag – the 1-bit non-negative integer value specifying a if the values of len_mask_2_x and len_mask_2_y are zero or greater than zero. B1 – the 16 bit value specifying the bias (additive component), used in section 14.1.1. nonLinear_enabled_U_flag – on/off switch for nonlinear filtering process of U component. nonLinear_enabled_V_flag – on/off switch for nonlinear filtering process of V component. nonlinearW – width of the weight tensor of the nonlinear filtering process. NonlinerW shall be multiple of 64. nonlinearH – height of the weight tensor of the nonlinear filtering process. NonlinerW shall be multiple of 64. 2. (6.5.1.1) Syntax table of hyper tensor z_stream() { Descriptor PCT } for (i =0; i < num_threads_z; ++i) { 3. (6.5. ) Qua y map n orma on decoder The input of this process is: – Codestream for quality_map; – Indicators for sigma tables which are decoded in picture header o quality_map_entropy_index[comp] for primary (comp==0) and secondary (comp==1) components. The output of this process is: – quality_map_delta[comp] is an array of size ^h4,w4^ with information used for deriving scaling factor (section 12.2) for primary (comp==0) and secondary (comp==1) components residual tensor. Here sizes h4, w4 are defined in Table 1. q_primary_sigma_Idx is a 1D array of size h4^w4 and each element in this array is equal to q_sigma_Idx[quality_map_entropy_index[comp]], where the q_sigma_Idx[k] is derived according to Table 4. q_comp[comp] is a 1D array of size h4^w4 and quality_map_delta[comp] is determined by: 42 F1250625PCT – quality_map_delta[comp][i, j] = q_comp[comp][ i^w4+j], for j = 0.. w4^1, and i = 0.. h4^1. Table 4 Determining q_sigma_Idx for quality map delta decoding k 0 1 2 3 4 5 6 7 q sigma Idx[k] 0 4 5 6 8 10 14 18 q_stream() { Descriptor SOQ u(16) 15. (6.5.3.1) Primary residual stream decoder The input of this process is: – Primary component residual codestream; – mask_skip[^^^, h4Y, w4Y] generated by Skip Mask generation (section 13.3.2) for primary component; – sigma_Idx_primary – tensor of size ^^^^, ℎସ^,^^ସ^൧ which is the output of Sigma quantization (section 10.8) for primary component. 43 F1250625PCT The output of this process is: – r_primary – one dimensional array ^^^^of residual tensor elements which is an input of decoder skip process (section 13.3.4) for primary component. Here sizes ^^^, ^^ସ^, ℎସ^ are defined in Table 2. A one dimensional array sigma_Idx_primary_1D is determined though following steps: ^ Set k = 0, ^ For ^^ ൌ 0..^^^ െ 1, ^ For ^^ ൌ 0.. ℎସ^ െ 1, ^ For ^^ ൌ 0..^^ସ^ െ 1, ^ if mask_skip[c,i, j] is equal to True, sigma_Idx_primary_1D[k] is set to equal to sigma_Idx_primary[c, i, j] and increase k by 1. 16. (6.5.3.1.1) Syntax table of primary residual tensor r_primary_stream() { Descriptor SOR (16) 17. (6.5.3.2) Secondary residual stream decoder The input of this process is: – Codestream for secondary component residual; – mask_skip[^^^ , h4UV, w4UV] generated by Skip Mask generation (section 13.3.2) for – sigma_Idx_secondary – tensor of size ^^^^, ℎସ^^ ,^^ସ^^^ which is the output of Sigma 44 F1250625PCT quantization (section 10.8) for secondary component. The output of this process is: – r_secondary – one dimensional array ^^^^ of residual tensor elements which is an input of decoder skip process (section 13.3.4) for secondary component. Here sizes ^^^, ℎସ^^ ,^^ସ^^ are defined in Table 2. A one dimensional array sigma_Idx_secondary_1D is determined though following steps: ^ Set k = 0, ^ For ^^ ൌ 0..^^^ െ 1, ^ For ^^ ൌ 0.. ℎସ^ െ 1, ^ For ^^ ൌ 0..^^ସ^ െ 1, ^ if mask_skip[c,i, j] is equal to True, sigma_Idx_secondary_1D[k] is set to equal to sigma_Idx_secondary[c, i, j] and increase k by 1. 8. (9.5.4.1) Syntax table of secondary residual tensor r_secondary_stream() { Descriptor SOR 16 9. (7.3) Hyper Scale Decoder Hyper scale decoder net is depicted in Fig.17. The input of hyper scale decoder is: ^ ^̂^^C, h6, w6^ reconstructed hyper tensor after me-tANS decoder (described in sections 9.6), 45 F1250625PCT ^ model parameters for Hyper Scale Decoder Net defined by modelIdx, all multiplier parameters in those models are 8-bits integer. The output of hyper scale decoder is standard deviation logarithm tensor I^^C, h6, w6^ with integer values in a range 0^ I^^(( N^ ^1) ^^ sigmaPrecision), where N^, sigmaPrecision are defined in section 15.4. Here h4, w4, h6, w6 are defined in the Table 1. All operation in scalable hyper decoder are integer, an accumulator in all computations is within 32 bits integer diapason, multipliers in model parameters are quantized to 8-bits integer. This guarantees bit-exact behaviour of this neural network module. Hyper scale decoder uses special neural network layers quantized convolutions. For each quantized convolution in the process the set of clipping values ^dk^ and de-scaling shifts parameters ^pk^ (1^k^3) are specified for each channel. All clipping values in quantized convolutions are dk=27, so the input of each convolution in Hyper Scale Decoder is a tensor of 8-bit integer values. De-scaling shifts ^pk^ are part of the model. NOTE ^ the magnitude of integerized weights in quantized model doesn’t exceed ^^^^, shift and clipping value combination ensures the register of quantized convolution is within 32-bits. The sequence of operations in this module doesn’t depend on operation point indicator (^^^^^^^^^^), but model parameters are different for base and high operation point. Hyper scale decoder starts with quantized convolution kernel size 1^1, stride 1, followed by rectified linear unit. Then there is a stride 1 quantized convolution with kernel size 3^3, followed by rectified linear unit. The next stride 1 quantized convolution (kernel size 1^1) increases the number of channels to 16^C . It is followed by shuffle (stride 4), which brings number of channels back to ^^ . The cropping layer (to size h4, w4) ensures the size of output tensor is ^C, h4, w4^. The process concluded with abs operation. Model with weights are storied in electronic attachment to this document, in format specicied in section 4.8 and locates in model_<MID>/<COMP>/hyper_scale_decoder.onnx, where <MID> is an integer from 0 to 3, <COMP> is “primary” or “secondary”. 0. (8.2) Hyper Decoder The learning-based hyper decoder consists of two independent pipe-lines with identical neural network architecture, except the number of channels; C=Cp for primary and C=Cs for secondary component (Cp and Cs are defined in Table 1). The input of this process is: ^ ^̂^^C, h6, w6^ reconstructed hyper tensor after me-tANS decoder (described in sections 46 F1250625PCT 9.6), ^ model parameters for Hyper Decoder Net defined by pair modelIdx. The output of this process is: ^ ^^^^4^C, h5, w5^ is re-shuffled explicit prediction tensor(part of predicton tensor derived from explicitly signalled information). Hyper decoder process is depicted in Fig.18. Hyper decoder starts stride 1 convolution with kernel size 1 ൈ 1, followed by inverse convolution (stride 2, kernel size 4 ൈ 4), cropping layer (depth 6) and leacky rectified linear unit. Next step is stride 1 convolution with kernel size 3 ൈ 3, followed by leacky rectified linear unit. Number of channels keeps un-changed till this point (equal to number of channels ^^ of input tensor). Then a stride 1 convolution layer with kernel size 3 ൈ 3 increases number of channels to 4^^, and followed by leacky rectified linear unit. Hyper decoder concluded by stride 1 convolution with kernel size 1 ൈ 1 followed by leacky rectified linear unit. Model with weights are storied in electronic attachment to this document, in format specicied in section 4.8 and locates in model_<MID>/<COMP>/hyper_decoder.onnx, where <MID> is an integer from 0 to 3, <COMP> is “primary” or “secondary”. 1. (8.3) Latent tensor reconstruction process The input of this process is: ^ ^̂^^C, h4, w4^ reconstructed residual tensor, which is an output of SKIP Model process (13.3.4), ^ ^^^^4^C, h5, w5^ re-shuffled explicit prediction tensor, which is an output of Hyper Decoder (11.2), The output of this process is: ^ ^^^′ ^C, h4, w4^ reconstructed latent tensor. ^^^ is obtained from ^^^ and ^^^ though Multi-stage Context Modelling process (section 11.4). 2. (8.4) Multistage Context Modelling The input of this process is: ^ ^̂^^C, h4, w4^ re-shuffled reconstructed residual tensor, which is an output of SKIP Model process (13.3.4), ^ ^^^^4^C, h5, w5^ re-shuffled explicit prediction tensor, which is an output of Hyper Decoder (11.2), ^ Four MCMk , k=0,…3 models with parameters defined by pair (modelIdx, k). 47 F1250625PCT The output of this process is: ^ ^^^ ^4^C, h5, w5^ re-shuffled reconstructed latent tensor. The process consists of following steps: ^ split padding layer (2h5, 2w5) and down-shuffle for reconstructed resdiaul ^̂^^^^, ℎସ,^^ସ^ to ^^^^4^C, h5, w5^ re-shuffled residual tensor, ^ split ^^^^4^C, h5, w5^ and ^^^^4^C, h5, w5^ into four parts of size ^C, h5, w5^ (each part consists of C out of 4C channels): o for l=0,..,3, ^ for c=0,.., C-1, ^ for i=0,.., h5-1, o for j=0,.., w5-1, ^ ^^^ l=^^^^l^C+c, i, j ^, ^ ^^^ l=^^^^l^C+c, i, j ^, ^ For l=0,…,3, o MCM(l) process which ^ takes as an input ^ ^^^^^^,^^ ൌ 0, … , ^^ െ 1 previously reconstructed parts of re-shaped ^ ^^^^ – collocated part of reconstructed residual tensor, ^ ^^^ ^- collocated part of explicit prediction tensor, ^ outputs ^ produces ^^^ l. ^ up-shuffle followed by cropping layer (h4, w4) ^^^ ^4^^, ℎହ,^^ହ^ to ^^^′^^^, ℎସ,^^ସ^. ^ Multi-stage context modelling process is depicted in Fig.19. This is a recurrent process: later stages uses previously obtained elements of output tensor as an input. Data flow which corresponds usage of previous stages output is marked with arrows in. Models with weights are storied in electronic attachment to this document, in format specicied in section 4.8 and locates in model_<MID>/<COMP>/MCM/stage<N>.onnx, where <MID> is an integer from 0 to 3, <COMP> is “primary” or “secondary”, <N> is a number of MCM’s stage from 0 to 3. 3. (10.3) Skip Mode Skip Mode allows skip writing to / parsing from the bit-stream residual tensor elements which can be identified by encoder and decoder to be zeros. 48 F1250625PCT 4. (10.3.1) Cube Flag Generation The input of Cube Flag generation process is: – ℎସ latent space tensor height, – ^^ସ latent space tensor width, – cube_flag[NumCubeFlag] – array of cube flags parsed from Picture Header (section 9.3), where NumCubeFlag =4*(^ℎସ ^ 15^ ≫ 4) ^(^^^ସ ^ 15^ ≫ 4). The output of Cube Flag generation process is: – cube_flag_skip[^^,ℎସ,^^ସ]. The cube size is ^^ ൈ 16 ൈ 16. Each cube contains 4 cube flags which store in the 1D arrays ^^^^^^^^_^^^^^^^^. For each element ^^^, ^^, ^^), ^^^^^^^^_^^^^^^^^_^^^^^^^^^^^, ^^, ^^^ ൌ ^^^^^^^^_^^^^^^^^^^^ ∗ ℎ ∗ ൫^^%2 ^ 2 ∗ ^^^%2^൯ ^ ^^ ∙ ^^^ ≫ 4^ ^ ^^^ ^ ^^ ^ ^ ^^ ^ ^ ^^ ^ ^ ൌ ^^^ସ ^ 15^ ≫ 4. 5. (10.3.2) SKIP Mask Generation The input of SKIP Mask generation process: – log domain standard deviation tensor ^^ఙ′ [C,h4,w4] after Sigma Scale (10.4); – compIdx: 0 - primary (y) or 1 - secondary (uv); – cube_luma_flag [^^, ℎସ,^^ସ] if (compIdx ==0) or cube_chroma_flag [^^, ℎସ,^^ସ] if (compIdx == 1). The output of SKIP Mask generation process is: – mask_skip[C,h4,w4]. First cube_flag_skip tensor fo size [^^,ℎସ,^^ସ] generated as described in section 13.3.1, using (compIdx ==0)? cube_luma_flag: cube_chroma_flag as input. The following ordered steps are applied: – Each one of the sigma samples is compared with the threshold_skip which is equal to 382 for all model_id. The comparison is stored in a mask tensor ^^^^^^^^_^^^^^^^^ of size ^^^, ℎସ,^^ସ^. o ^^^ ^^^^^^^^^ if ^^ఙ′ ^^^, ^^, ^^^ ^ threshold_skip . Then ∪ (! cube_flag_skip^^^, ^^, ^^^). 6. (11.2) Adaptive linear filter 49 F1250625PCT This section details the primary component guided adaptive linear filter process for secondary component. This process provides enhancement of secondary components (colour information planes) of image utilizing information from primary component. This process is enabled if EFE_linear_filter_enabled_flag is true. The input of this process is: ^ ^^^^ ^1,^^^,^^^^ (output of synthesis transfor for primary component), ^ ^^^^^ ^2,^^^^^^,^^^^^^^ (output of synthesis transfor for secondary component). The output of this process is: ^ enhanced secondary component ^^^′ ^^ ^ 2,^^^^,^^^^ ^ which goes to the ICCI filter block (section 14.1) and ^^^′′ ^2,^^^^,^^^ the non-linear filter block (section ^^ ^ 14.3). If is equal to 1 the following ordered steps are performed: ^ The parsing process according to parsing table in section 9.4.1 is invoked to obtain ^^^^, ^^^^, ^^ସ^, ^^ସ^ and ^^^. ^ Tiling process is as described in section 14.2.2 is invoked with parsed syntax elements as inputs and Tile1 tensor as output. ^ Parameter update process as specified in section 14.2.1 is invoked with ^^ସ^ and ^^^^as inputs and modified ^^ସ^ and ^^^^ as outputs; ^ ^^^[2] is vector is subtracted channelwise from ^^^′ ^^[2, ^^^^^^,^^^^^^]; ^ ^^ ^2 ∙ ^^^^^^^^^^௩^^ ∙ ^^^^^^^^^^^^^,ுା^ ଶ ,^ା^ ଶ ^ and ^^ ^4,ுା^ ଶ , ^ା^ ଶ ^ are set equal to ^ For x = 0.. (W+1)/2-1, y = 0…(H+1)/2-1, k = 0..1, i=0…^^௩^^-1, j=0…^^^^^-1 the following is performed: ^ ^^ℎ ൌ 4^^ ^ 2^^ ^ ^^, ^ ^^ℎ^௨௧ ൌ ^^ ∙ ^^௩^^ ∙ ^^^^^ ^ ^^^^^ ∙ ^^ ^ ^^, ^ ∙ ^^^%^^^^^^^^^^௩^^^ ^ ^^^%^^^^^^^^^^^^^^, ^ ^^^ௗ௫ ൌ ^^^^^^^^1^^^, ^^, ^^^, ^ ^^′^^^ℎ^௨௧,^^, ^^^ ൌ ^^^^^ℎ^^,^^, ^^^ ⋆ ^^ସ^^^^^ௗ௫ , ^^ℎ^ ^ ^^^2^^ ^ ^^,^^, ^^^ ⋆ ^^ସ^^^^^ௗ௫,^^^ ^ ^^^[k], ^ ^^′^^^ℎ^௨௧,^^, ^^^ ൌ ^^^^^ℎ^^,^^, ^^^ ⋆ ^^^^^0, ^^ℎ^+ ^^^2^^ ^ ^^,^^, ^^^ ⋆ ^^^^^0, ^^^ ^ ^^^[k]. ^^^′ ′ ^^ ^2,^^^^ ,^^^^^ and ^^^′′^^^2,^^^^ ,^^^^^ are set equal to pixelshuffle^^^ , ^^௩^^ , ^^^^^^ and 50 F1250625PCT shuffle^^^′′, ^^௩^^ , ^^^^^^ respectively. 27. (11.4) Luma Edge Filtering (LEF) filter This section details the Luma Edge Filtering(LEF) process. The input of this process is: – ^^^′ ^1,^^ ,^^ ^ (the output of ICCI process 14.1); ^ ^ ^ – log domain standard deviation tensor ^^ఙ^′ [C,h4,w4] for primary component after Sigma Scale (10.4). The output of this process is ^ ^ ^^^^^,^^^^. Process 10.5 Conversion of a standard deviation from logarithmic to linear scale is invoked with ^^′ ఙ^ as input and ^^′ ^ as output. If LEF_enabled flag is false, final reconstructed primary ^^^′ ^1,^^^,^^ ^ will be unchanged. ^ ^ If LEF_enabled flag is true then the process specified in section 9.4 is performed. 28. (11.4.1) LEF general process LEF process receives primary variance tensor ^^^^^^^^, ℎସ,^,^^ସ,^൧ and reconstructed primary ^^^′ ^ ^1,^^,^^^ as inputs. The output of LEF process is edge enhanced modified ^^^′ ^ ^ 1,^^, ^^ ^ . The following ordered steps are performed: ^ Picture header parsing process is invoked as described in section 9.4to obtain ^^^^^^^^^^^^^^^^^^^^^^^^; ^ ^^ℎ^^^^^^ is set equal to ^^^^^^^^^^^^^^^^^^_^^ℎ^^^^^^^^^^_^^^^^^_^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^; ^ The ^^^^ ^_^^^^^^^ ^1,^^, W^ tensor is obtained as follows, for ^^ ൌ 0..^^ െ 1, ^^ ൌ 0.. ^^ െ 1; ^ ^^^ ^ ^0, ^^, ^^^ ^^^^^^^ℎ^^^^^^, ^^^^^^^^^^^^^ ∙ ^^^^, ^^^^^^^^^^^^^ ∙ ^^^^^; ^ ^ intensity[4]=luma_edge_filter_intensity_list [targetBppIdx]; ^ ^^^′ ^ ൌ ^^^′ ^ ൊ 255; ^ ^^^^^௨^ ൌ ^^^′ ^ ⋆ ^^^^^^^^^^^^; ൊ 8190; 51 F1250625PCT 0, ^^^ ^ ì ^^^ೌ^^^^ ^0, ^^, ^^^ ^ ^^ℎ^^^0^ ï 1, ^^ℎ^^^0^ ^ ^^^ ^ ^^^ೌ^^^^ ^0, ^^, ^^^ ^ ^^ℎ^^^1^ ^ ^^^^^^^^ ^^^ ^ In this when the indices of a tensor exceeds the tensor boundaries. The as follows: 1 ൊ 13 1 ൊ 13 1 ൊ 13 ^^^^^^^^^^^^ ൌ ^ 1 ൊ 13 5 ൊ 13 1 ൊ 13൩ 1 ൊ 13 1 ൊ 13 1 ൊ 13 The ^^^^^^^^^^^^^^^^^^_^^ℎ^^^^^^^^^^_^^^^^^_^^^^^^^^^^ is defined as follows: ^^^^^^^^^^^^^^^^^^_^^ℎ^^^^^^^^^^_^^^^^^_^^^^^^^^^^ ൌ ^23,23,65,65,65^ The ^^^^^^^^_^^^^^^^^_^^^^^^^^^^^^_^^^^^^^^^^^^^^^^^^_^^^^^^^^ is as follows: ^^^^^^^^_^^^^^^^^_^^^^^^^^^^^^_^^^^^^^^^^^^^^^^^^_^^^^^^^^ ^^^^^^^^ Y=0 Y=1 Y=2 Y=3 X 1 11 12 140 6 7 4 The ^^^^^^^^_^ _ _ _ ℎ _ _ Y=0 Y=1 Y=2 0 0 0 0 2.14 Model header model_header(comp) { Descriptor 52 F1250625PCT tile_header(comp) independent_beta_uv is a flag (false/true), independent_beta_uv equal to true indicates that the b t. If independent_beta_uv equal to false then beta displacement parameter for primary and secondary components are the same. beta_displacement_log_plus_2048[comp] minus 2048 is a – parameter indicating a displacement between the rate control parameter beta selected by encoder for comp component and the reference rate control parameter beta associated with the index of the used model (model_id). The displacement is in logarithmic scale. betaDisplacementLog[comp] = beta_displacement_log_plus_2048[comp] – 211 NOTE – the reference rate control parameter beta (β), mentioned here, is the parameter used in the model training to control the ratio between the bitrate and distortion. “The model” here means the model associated with the index of the used model (model_id). When syntax element beta_displacement_log_plus_2048[1] is not present (independent_beta_uv is equal to 0), betaDisplacementLog[1] = betaDisplacementLog[0]. 2.15 Variable Rate Support 2.15.1 General This section 2.15 describes operations which are used for variable rate coding. 2.15.2 Control parameters and gain tensor derivation In total four models (defined by model_id =0…3) are trained for the different ranges of quality. Pre-trained model is selected base on model_id the syntax element coded at Picture header. The learnable model includes the reference forward gain vector for each component comp in logarithmic scale mlog^model_id^^comp^^C^, comprising 12-bit signed values. NOTE – Each model_id is associated with the reference r ate control parameter ^^, which was used in the model training to control the ratio between the bitrate and distortion. The forward gain tensor m is used at encoder side in Gain Unit. Forward gain tensor in logarithmic scale ^^^^^ is used in Sigma scale. The inverse gain tensor ^^ି^ is used at decoder side in Inverse Gain Unit. All three forward ^^ , inverse ^^ି^ and logarithmics domain ^^^^^ gain tensors have size ^^^, ℎସ,^^ସ^ equal to the size of the residual tensor for the component. Component index comp is equal to 0 for the primary color component and 1 for the secondary component. The input of gain tensor derivation process is: ^ 12-bits signed variable betaDisplacementLog[comp]; 53 F1250625PCT ^ tensor with 3-bit elements ^^^^^^^^^^^^^^_^^^^^^_^^^^^^^^^^^^^^^^^^^^^ℎସ,^^ସ^; ^ index model_id which indicates pretrained model to be used. The output of gain tensor derivation process is: ^ forward gain tensor in logarithmic scale ^^^^^ ^^^, ℎସ,^^ସ^; ^ forward gain tensor ^^ ^^^, ℎସ,^^ସ^; ^ inverse gain tensor ^^ି^ ^^^, ℎସ,^^ସ^. Sizes of all tensors ^^, ℎସ,^^ସ for primary and secondary components are defined in Table 1. Forward gain tensor in logarithmic scale ^^^^^ is computed as following: For 0 ^ ^^ ^ ^^; 0 ^ ^^ ^ ℎସ; 0 ^ ^^ ^ ^^ସ: ^^^^^^^^, ^^, ^^^ ൌ ^^^^^^^^^^^^^^^^^ௗ^^^^^^^^^^^^^^^ ^ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ^ ^^^^^^^^3^^^^^^^^^^^^^^^^^_^^^^^^^^^, ^^^ ^ where quality_map[i,j] is equal to 0 when gain_3D_enable_flag[comp] is equal to 0 and, otherwise, is equal to ^^^^^^^^^െ8,8,^^^^^^^^^^^^^^_^^^^^^_^^^^^^^^^^^, ^^^ ^ ^^^^^^^^^^^^^^_^^^^^^_^^^^^^^^^^^comp^ ^^^, ^^^^, െ16 ^ ^^^^^^^^^^^^^^_^^^^^^_^^^^^^^^^^^^^^^^^^^^ ^^^, ^^^ ^ 16 o ^^^^^^^^^^^^^^_^^^^^^_^^^^^^^^^^^, ^^^ ൌ ^^ ^^ 0. by ^^^^^^^^3^^ in Table 5. There are in total 17 quality levels. Table 5 Additional beta-displacement for local quality control ^^^^^^^^^^^^^^_^^^^^^ 0 1 2 3 4 5 6 7 8 6 ^^^^^^^^^^^^^^_^^^^^^ -8 -7 -6 -5 -4 -3 -2 -1 1 8 NOTE – if gain_3D_enable_flag is equal to 0, then mlog can be calculated as a vector with size C from the vector mlog [model_id] with the similar size and betaDisplacementLog^comp^. Later on this vector mlog can be expanded to the size C×h4×w4 filling the whole channel with the value mlog[c]. NOTE – The value of gain3D is computed from the factor according to the following 54 F1250625PCT equations: – ^^^^^^^^^^^^^^^^ ൌ ^ ୪୬^ఙ^ೌ^^ି୪୬ ^ఙ^^^^ ^ ே^ି^ – ^ ^^^^^^^^^^^^^^ ^^ି^ are computed as following: – ே^ି^ ସ are defined in section 2.15.3 Gain Unit This process is applied at encoder side only and so it is non-normative process, described here to simplify explanation of normative inverse gain unit and sigma scale processes. The input of this process: ^ forward gain tensor ^^, ^ residual tensor ^^ of size ^^^,ℎସ,^^ସ^ , which is the result of subtraction mean value from latent tensor ^^ ൌ ^^ െ ^^. The output of this process: ^ scaled by forward gain tensor residual tensor r’ of size ^^^, ℎସ,^^ସ^. On encoder side a gain unit is placed after a residual calculation. An output scaled residual tensor is equal to the input tensor multiplied by gain tensor in a channel dimension: ^^′^^^, ^^, ^^^ ൌ ^^^^^, ^^, ^^^ ∙ ^^^^^, ^^, ^^^; 0 ^ ^^ ^ ℎସ, 0 ^ ^^ ^ ^^ସ, 0 ^ ^^ ^ ^^. All elements of the tensor is the same channel are scaled by the same multiplier. 2.15.4 Inverse Gain Unit This process is applied at decoder side and so normative. The input of this process: ^ inverse gain tensor ^^ି^, ^ reconstructed residual tensor ^̂^ of size ^^^, ℎସ,^^ସ^ , which is the output of decoder-size SKIP process. The output of this process: ^ scaled by inverse gain tensor residual tensor ^̂^′ of size ^^^, ℎସ,^^ସ^. 55 F1250625PCT On decoder side an inverse gain unit is placed after entropy decoder and takes reconstructed residual as an input . The output is reconstructed residual tensor multiplied by inverse gain tensor: ^̂^′^^^, ^^, ^^^ ൌ ^^ି^^^^, ^^, ^^^ ∙ ^̂^^^^, ^^, ^^^; 0 ^ ^^ ^ ℎସ, 0 ^ ^^ ^ ^^ସ, 0 ^ ^^ ^ ^^. All elements of the tensor in the same channel are scaled by the same multiplier. 2.15.5 Sigma Scale This process is applied both at encoder and decoder sides, and so normative. The input of this process are: ^ forward gain tensor in logarithmic scale mlog, ^ standard deviation logarithm tensor I^ of size ^C, h4, w4^ , which is and output of Hyper Scale Decoder. The output of this process is: ^ scaled by forward gain tensor standard deviation logarithm tensor I’^ of size ^C, h4, w4^ which goes to the adaptive sigma scale. Sigma scale modifies standard deviation logarithm tensor I^^C, h4, w4^ as following: I’^^c, i, j^ = mlog^c, i, j^+ I’^^c, i, j^, i=0,..,h4-1; j=0,…, w4-1, c=0,…,C-1. 2.15.6 Sigma quantization The input of this process is: – I’’^^ C,h4,w4^ which is an output of Adaptive Sigma Scale. The output of this process is: – sigma_Idx[C,h4,w4] which is further used in Entropy Decoder for ^̂^ . The process is as follows: For c=0,…C-1 , i=0,…, ℎସ െ1, j=0,…, ^^ସ െ1: sigma_Idx [c,i,j] = clip(I’’^+2sigmaPrecision^1)^^ sigmaPrecision, 0, N^^1), While computing CDF table for sigma_Idx=i, 0 ^i^ N^ =^^ it is assumed that: ^min=exp((ln(^max)- ln )^i/( N^^1)+ln(^min)). 3. Problems 1. On signaling skip mode syntax elements In an existing design, the following problems exist on signaling skip mode syntax elements to support regional accessibility: Fig.22 illustrates an MCM model to obtain cube_flag, which will be encoded to the bitstreams. As shown in Fig. 22, the ^^_^^^^^^ denotes the latent tensor after analysis transform, and ^^^^^^^^_^^^^ is the final dequantized residual. In MCM, the input ^^_^^^^^^ will 56 F1250625PCT first down-shuffled into 4 sub-tensors, whose spatial size (width or height) is half of the input. The prediction process is proceeding one by one sub-tensor, where for each sub-tensor, e.g., ^^0, max pooling across spatial size of ^^ ൈ ^^ will be performed. In one example, ^^=8. ^^^^^^^^_^^^^^^^^ is used to derive ^^^^^^^^_^^^^^^^^^^_^^^^^^^^ through up-shuffle operation, which will be finally used to guide the entropy coder in encoding and decoding latent residual ^^^^^^^^_^^. When the whole is fed into the analysis transform, the MCM will obtain the cube_flag for the whole image as a single tensor. However, when the image is split into multiple regions and in the MCM each time one region is processed. The problem arises since there is overlapping in the latent tensor and inside ^^^^^^_^^^^^^^^_^^^^^^^^ module the max polling operation is applied. 2. On variable rate support In image and video coding, it is sometimes necessary to generate a bitstream that is close to a target bitrate (e.g. variable rate coding). However, achieving this requires additional processing, and hence increases the complexity. In majority of the applications, rate matching (variable rate support) is not used, as the application has a target quality, and not a target bitrate. Therefore, the supporting variable rate coding increases the burden on majority of the decoders without any benefit. In order to achieve variable rate coding, one common method is to scale parameters that are encoded in the bitstream before quantization, and applying an inverse scaling after quantization. In an existing design, this is achieved by variable rate support as described in the above section 2.15. In the above-mentioned existing design, however, it is not possible to turn on or off the variable rate functionality. An encoder always needs to signal at least one parameter (e.g. independent_beta_uv and/or beta_displacement_log_ plus_2048[comp]), which then need to be used in the following functions: - betaDisplacementLog[comp] = beta_displacement_log_plus_2048[comp] – 211, - ^^^^^^^^, ^^, ^^^ ൌ ^^^^^^^^^^^^^^^^^ௗ^^^^^^^^^^^^^^^ ^ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ^ - ... As one can see, series of operations need to be performed by all decoders, although the variable rate coding is used only by certain applications. 4. Detailed Solutions 57 F1250625PCT To solve the above-described problems, methods as summarized below are disclosed. The solutions should be considered as examples to explain the general concepts and should not be interpreted in a narrow way. Furthermore, these solutions can be applied individually or combined in any manner. The proposed solution(s) primarily includes three aspects: a) skip mode signaling; 2) bitstream structure; 3) variable rate support. 4.1 Skip mode tool header signaling In the above Section 8 (6.3.1.1) Model header, skip_mode_header() is described, wherein cube_flag for the whole image is encoded to the bistreams. In this solution, it is proposed to reorganize the cube_flag. The image is divided into regions, thereby, multiple region_flag tensors will be derived from these regions, forming a cube_flag list. When define the number of region rows as NumHorSplits and number of region columns as NumVerSplits, the number of regions is N = NumHorSplits * NumVerSplits. In this cube_flag_list, there will be ^^ tensor, each corresponds to one of the regions. The encoder will encode these N tensors sequentially. 1) In one example, cube_flag tensor is reorganized, wherein each region produces a distinct cube_flag. The encoder encodes ^^ cube_flag tensors sequentially, where ^^ indicates the number of regions. 2) Alternatively, in one example, a list of cube_flag_header(), for example, cube_flag_list_header() is added for skip mode. For purpose of illustration, the following table shows modified skip mode signaling, wherein encoding cube_flag_header() is repeated for N times, where N denotes the number of regions. Most relevant parts that have been added or modified are shown by using bolded words (e.g., this format indicates added text), and some of the deleted parts are shown by using words in italics between double curly brackets (e.g., {{this format indicates deleted text}}). There may be some other changes that are editorial in nature and thus not highlighted. It should be understood that only markings in this table are intended to represent changes. 58 F1250625PCT For idx in 0,1, ... ,num-regions - 1 { Bitstream structure 1) Fig.23 illustrates the bitstream structure of q-stream, RY-stream and Ruv-stream are split based on regions, while z-stream as a single without split based on regions. As detailed in Fig. 23, it is proposed that to split q-stream, RY-stream and RUV-stream in the bitstreams based on the regions, while keep z-stream as a single stream. 2) Alternatively, Fig.24 illustrates the bitstream structure of q-stream, RY-stream, RUV-stream and z-stream are all split based on regions. as detailed in Fig.24, it is proposed that to split q-stream, RY-stream, RUV-stream and z-stream based on the regions. 3) Fig. 25 illustrates an example bitstream structure, where Z-stream is not split, number of threads are signaled between tools header and SOZ. When z-stream is not split, alternative to Fig. 23, the bitstream structure shown in Fig. 25 may apply, wherein the number of threads of z, q, r_primary and r_secondary are right after tools header and before SOZ. 4) Fig. 26 illustrates an example bitstream structure, where Z-stream is not split. Number of thread z is signaled after tools header and before SOZ, while number threads q, primary, and secondary are after z-stream and before region data. When z-stream is not split, 59 F1250625PCT alternative to Fig. 23 or Fig. 25, the bitstream structure shown in Fig. 26 may apply, wherein the number of threads of z is proceeding SOZ, and number of threads of q, r_primary and r_secondary are proceeding region data. When z-stream is not split (as shown in Fig. 23), there are some alternative bitstream structures. Fig.25 illustrates an example bitstream structure, where Z-stream is not split, number of threads are signaled between tools header and SOZ. Fig. 26 illustrates an example bitstream structure, where Z-stream is not split. Number of thread z is signaled after tools header and before SOZ, while number threads q, primary, and secondary are after z-stream and before region data. 4.3 On variable rate support An indication is included in (or decoded from) a bitstream to enable or disable variable rate support. ^ The indicator might control at least one of the following: o if variable rate coding is enabled or disabled. o The indicator might indicate if variable rate coding is enabled or disabled. o The indicator might indicate if a residual parameter or a sigma parameter is modified based on a default value or a value obtained from a bitstream. ^ There might be multiple default values. For example, the default value might be obtained based on an index (e.g. a model index). ^ In one example if the indicator is false, the modification of sigma and/or residual parameters are performed based on a default parameter. And if the indicator is true, the modification is performed based on a default parameter and a parameter obtained from a bitstream. ^ A specific implementation might be as follows: ^ ^^^^^^^^, ^^, ^^^ ൌ ^^^^^^^^^^^^^^^^^ௗ^^^^^^^^^^^^^^^ ^ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^, wherein the ^^^^^^^^^^^^^^^^^ௗ^^^^^^^^^^^^^^^ is the default parameter, and the value of ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ is equal to zero if the indicator false, and the value of ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ might be different from zero when the indicator is true. Sigma and/or residual parameters might be modified based on ^^^^^^^^, ^^, ^^^. ^ The indicator might be an enabled flag. o The indicator might indicate if additional parameters that are associated with variable rate coding are included in the bitstream. ^ The additional parameter might be independent_beta_uv or 60 F1250625PCT beta_displacement_log_ plus_2048. o The indicator might be beta_displacement_present_flag, or beta_displacement_inferred_flag or variable_rate_coding_enabled flag. ^ The indicator might indicate if the value of a parameter is inferred. o The value of the parameter might be inferred to be equal to 0. Or it might be inferred to be equal to 2048. o The value of beta_displacement_log_ plus_2048 might be inferred to be equal to zero. o The value of beta_displacement_log_ plus_2048 might be inferred to be equal to 2048. o The value of betaDisplacementLog might be inferred to be equal to 0. ^ The indicator might be included in a header. o The indicator might be included in a picture header. ^ The indicator might be included right before the model header inside the picture header. o The indicator might be included in a model header. ^ The indicator might be the first syntax element of the model header. General aspects 5) Additional operations may be applied to or with the proposed method. a) A syntax element (a.k.a. an indication) disclosed above may be binarized as a flag, a fixed length code, an EG(x) code, a unary code, a truncated unary code, a truncated binary code, etc. It can be signed or unsigned. b) A syntax element representing a coding tool or a coding method may not be signalled and implicitly determined to be unused, if the coding tool or the coding method is regarded as not applicable or cannot be used. c) A syntax element disclosed above may be coded with at least one context model. Or it may be bypass coded. d) A syntax element disclosed above may be signaled in a conditional way. e) A syntax element disclosed above may be signaled at block level/ sequence level/group of pictures level/picture level/slice level/tile group level. f) Whether to and/or how to apply the disclosed methods above may be signalled at block level/ sequence level/group of pictures level/picture level/slice level/tile group level. g) Whether to and/or how to apply the disclosed methods above may be dependent on coded information, such as colour format, colour component, slice/picture type. 6) The proposed methods may be applied to other image/video compression solutions with 61 F1250625PCT NN-based coding tools involved. 5. Embodiments 5.1 Skip mode tool header As shown in the following table, the cube_flag_list includes NumVerSplits * NumHorSplits tensors, wherein each tensor corresponds to each region. The encoder will encode cube_flag for each region one by one. RC[d][i][j][n] will be used to determine the region coordinates and width/height. Most relevant parts that have been added or modified are shown by using bolded words (e.g., this format indicates added text), and some of the deleted parts are shown by using words in italics between double curly brackets (e.g., {{this format indicates deleted text}}). There may be some other changes that are editorial in nature and thus not highlighted. It should be understood that only markings in this table are intended to represent changes. for (y=0; y< NumHorSplits; y++) { for (x=0 x<NumVerS lits x++) { r 62 F1250625PCT for( j = 0; j < NumCubeInGroup; j++ ) { if( comp == 0) The input of SKIP Mask generation process: – log domain standard deviation tensor ^^ఙ′ [C,h4,w4] after Sigma Scale (10.4); – compIdx: 0 - primary (y) or 1 - secondary (uv); – cube_luma_flag_list[compIdx]. The output of SKIP Mask generation process is: – mask_skip[C, h4, w4]. First cube_flag_skip tensor for size [^^, ℎସ,^^ସ] generated as described in section 13.3.1, using (compIdx ==0)? cube_luma_flag: cube_chroma_flag as input. The following ordered steps are applied: for y=0,…,numHorSplits-1; and x=0,…, verHorSplits, mask_skip_tile = mask_skip_list[verHorSplits*y+x], – Each one of the sigma samples is compared with the threshold_skip which is equal to 382 for all model_id. The comparison is stored in a mask tensor ^^^^^^^^_^^^^^^^^ of size ^^^,ℎସ,^^ସ^. o ^^^ ^^^^^^^^^ if ^^ఙ′ ^^^, ^^, ^^^ ^ threshold_skip . o ^^^^^^^^_^^^^^^^^_^^^^^^^^^^^, ^^, ^^^ ∪ (! cube_flag_list[verHorSplits*y+x]^^^, ^^, ^^^). o Crop out boundary padding: ^ y_start = 0 if y==0 else 8; x_start=0 if x==0 else 8, ^ y_end = RC[4][y][x][4] if y==numHorSplits-1 else RC[4][y][x][4]-8, 63 F1250625PCT ^ x_end = RC[4][y][x][5] if x==numVerSplits-1 else RC[4][y][x][5]-8, ^ mask_skip_tile=mask_skip_tile[:][:][y_start:y_end][x_start:x_end]. mask_skip[:,:,RC[d][y][x][1]- RC[d][y][x][0], RC[d][y][x][3]: RC[d][y][x][2]] = mask_skip_tile. 5.3 SKIP Mask generation The input of SKIP Mask generation process: – log domain standard deviation tensor ^^ఙ′ [C,h4,w4] after Sigma Scale; – compIdx: 0 - primary (y) or 1 - secondary (uv); – cube_luma_flag_list[compIdx]. The output of SKIP Mask generation process is: – mask_skip[C, h4, w4]. First cube_flag_skip tensor for size [^^, ℎସ,^^ସ] generated as described in section 13.3.1, using (compIdx ==0)? cube_luma_flag: cube_chroma_flag as input. The following ordered steps are applied: for y=0,…,numHorSplits-1; and x=0,…, verHorSplits, mask_skip_tile = mask_skip_list[verHorSplits*y+x], – Each one of the sigma samples is compared with the threshold_skip which is equal to 382 for all model_id. The comparison is stored in a mask tensor ^^^^^^^^_^^^^^^^^ of size ^^^, ℎସ,^^ସ^. o ^^^^^^^^_^^^^^^^^_^^^^^^^^^^^, ^^, ^^^ ൌ ^^^^^^^^^ if ^^ఙ′ ^^^, ^^, ^^^ ^ threshold_skip . o ^^^^^^^^_^^^^^^^^_^^^^^^^^^^^, ^^, ^^^ ∪ (! cube_flag_list[verHorSplits*y+x]^^^, ^^, ^^^). o Crop out boundary padding: ^ y_start = 0 if y==0 else 8; x_start=0 if x==0 else 8, ^ y_end = RC[4][y][x][4] if y==numHorSplits-1 else RC[4][y][x][4]-8, ^ x_end = RC[4][y][x][5] if x==numVerSplits-1 else RC[4][y][x][5]-8, ^ mask_skip_tile=mask_skip_tile[:][:][y_start:y_end][x_start:x_end]. mask_skip[:,:,RC[d][y][x][1]- RC[d][y][x][0], RC[d][y][x][3]: RC[d][y][x][2]] = mask_skip_tile. 5.4 (13.3) Skip Mode Skip Mode allows skip writing to / parsing from the bit-stream residual tensor elements which 64 F1250625PCT can be identified by encoder and decoder to be zeros. 5.4.1 (13.3.1) Cube Flag Generation The following steps are repeatedly executed for numVerSplits * numHorSplits, i.e., the number of regions. The cube_flag is saved into a list to be encoded. The input of Cube Flag generation process is: – ℎସ latent space tensor height, – ^^ସ latent space tensor width, – for (y=0; y<numHorSplits; y++) { for (x=0; x<numVerSplits; x++) { ℎ௬௫ ൌ ^^^^^4^^^^^^^^^^4^ ^ ^0 ^^^^ ^^ ൌൌ 0 ^^^^^^^^ 8^ ^ ^0 ^^^^ ^^ ൌൌ ^^^^^^^^^^^^^^^^^^^^^^^^ െ 1 ^^^^^^^^ 0^ ^^௬௫ ൌ ^^^^^4^^^^^^^^^^5^ ^ ^0 ^^^^ ^^ ൌൌ 0 ^^^^^^^^ 8^ ^ ^^^^ ^^ ^^^^^^^^^^^^^^^^^^^^^^^^ െ 1 ^^^^^^^^ 0^ cube_flag , = 7^ ≫ 3) ^(^^^௬௫ ^ 7^ ≫ 3) cube_flag_list.append(cube_flag) } } The output of Cube Flag generation process is: – cube_flag_skip_list[numHorSplits * numVerSplits]. The cube size is ^^ ൈ 8 ൈ 8, the 1D arrays ^^^^^^^^_^^^^^^^^ is represent a 2D array with height ℎ and width ^^ in raster scan order (an example is shown in Fig. 27), where ℎ ൌ ^ℎସ ^ 7^ ≫ 3,^^ ൌ ^^^ସ ^ 7^ ≫ 3. For each element ^^^, ^^, ^^), ^^^^^^^^_^^^^^^^^_^^^^^^^^^^^, ^^, ^^^ ൌ ^^^^^^^^_^^^^^^^^^^^ ∙ ^^^ ≫ 3^ ^ ^^^ ≫ 3^^. 5.4.2 (13.3.2) SKIP Mask Generation The input of SKIP Mask generation process: – log domain standard deviation tensor ^^ఙ′ [C,h4,w4] after Sigma Scale (10.4); – compIdx: 0 - primary (y) or 1 - secondary (uv); – cube_luma_flag_list[compIdx]; The output of SKIP Mask generation process is: mask_skip[C, h4, w4]. First cube_flag_skip tensor for size [^^, ℎସ,^^ସ] generated as described in section 13.3.1, using (compIdx ==0)? cube_luma_flag: cube_chroma_flag as input. 65 F1250625PCT The following ordered steps are applied: for y=0,…,numHorSplits-1; and x=0,…, verHorSplits, mask_skip_tile = mask_skip_list[verHorSplits*y+x], – Each one of the sigma samples is compared with the threshold_skip which is equal to 382 for all model_id. The comparison is stored in a mask tensor ^^^^^^^^_^^^^^^^^ of size ^^^, ℎସ,^^ସ^. o ^^^^^^^^_^^^^^^^^_^^^^^^^^^^^, ^^, ^^^ ൌ ^^^^^^^^^ if ^^ఙ′ ^^^, ^^, ^^^ ^ threshold_skip . ^^^^^^^^^^ ^^, ^^^ ∪ (! . o out = if x==0 else 8, ^ y_end = RC[4][y][x][4] if y==numHorSplits-1 else RC[4][y][x][4]-8, ^ x_end = RC[4][y][x][5] if x==numVerSplits-1 else RC[4][y][x][5]-8, ^ mask_skip_tile=mask_skip_tile[:][:][y_start:y_end][x_start:x_end], ^ mask_skip[:,:,RC[d][y][x][1]- RC[d][y][x][0], RC[d][y][x][3]: RC[d][y][x][2]] = mask_skip_tile. 5.5 On variable rate support 5.5.1 Decoder/encoder solution 1 Performing a conversion using a neural network between a bitstream and a reconstructed image or video based on an indicator in the bitstream, wherein, ^ If the indicator indicates a positive action, variable rate coding is used. ^ Otherwise variable rate coding is not used. 5.5.2 Decoder/encoder solution 2 Performing a conversion using a neural network between a bitstream and a reconstructed image or video based on an indicator in the bitstream, wherein, ^ If the indicator indicates a positive action, a value obtained from a bitstream is used in variable rate coding process. ^ Otherwise a default value is used in variable rate coding process. 5.5.3 Decoder/encoder solution 3 Performing a conversion using a neural network between a bitstream and a reconstructed image or video based on an indicator in the bitstream, wherein, ^ If the indicator indicates a positive action, a value obtained from a bitstream is used in variable rate coding process. 66 F1250625PCT ^ Otherwise a default value from a set of default values is used in variable rate coding process. 5.5.4 According to the above solution 3, the default value is selected from a set of default values based on an index indicated in the bitstream. 6. Example implementation of the proposed solutions Some relevant parts that have been added or modified are shown by using bolded words (e.g., this format indicates added text), and some of the deleted parts are shown by using words in italics between double curly brackets (e.g., {{this format indicates deleted text}}). There may be some other changes that are not highlighted. It should be understood that only markings in this section are intended to emphasize at least part of proposed changes. Model header model_header(comp) { Descriptor v ri bl r t din n bl d[ m ] (1) variable_rate_coding_enabled[comp] is a flag (false/true), indicating if variable rate coding is enabled or disabled. variable_rate_coding_enabled[comp] is equal to false, beta_displacement_log_ plus_2048[comp] is inferred to be equal to 2048. independent_beta_uv is a flag (false/true), independent_beta_uv equal to true indicates that the beta displacement parameter for primary and secondary components are different. If independent_beta_uv equal to false then beta displacement parameter for primary and secondary components are the same. beta_displacement_log_plus_2048[comp] minus 2048 is a – parameter indicating a displacement between the rate control parameter beta selected by encoder for comp component and the reference rate control parameter beta associated with the index of the used model (model_id). The 67 F1250625PCT displacement is in logarithmic scale. betaDisplacementLog[comp] = beta_displacement_log_plus_2048[comp] – 211 NOTE – the reference rate control parameter beta (β), mentioned here, is the parameter used in the model training to control the ratio between the bitrate and distortion. “The model” here means the model associated with the index of the used model (model_id). When syntax element beta_displacement_log_plus_2048[1] is not present (independent_beta_uv is equal to 0), betaDisplacementLog[1] = betaDisplacementLog[0]. 7. Another example implementation of the proposed solutions Some relevant parts that have been added or modified are shown by using bolded words (e.g., this format indicates added text), and some of the deleted parts are shown by using words in italics between double curly brackets (e.g., {{this format indicates deleted text}}). There may be some other changes that are not highlighted. It should be understood that only markings in this section are intended to emphasize at least part of proposed changes. picture_header( ) { Descriptor 68 F1250625PCT Model header model_header(comp) { Descriptor if (variable_rate_coding_enabled){ v _ _ _ , ng is enabled or disabled. variable_rate_coding_enabled is equal to false, the values of independent_beta_uv and beta_displacement_log_ plus_2048[0] and beta_displacement_log_ plus_2048[1] are inferred to be equal to false, 2048 and 2048 respectively. independent_beta_uv is a flag (false/true), independent_beta_uv equal to true indicates that the beta displacement parameter for primary and secondary components are different. If independent_beta_uv equal to false then beta displacement parameter for primary and secondary components are the same. beta_displacement_log_plus_2048[comp] minus 2048 is a – parameter indicating a displacement between the rate control parameter beta selected by encoder for comp component and the reference rate control parameter beta associated with the index of the used model (model_id). The displacement is in logarithmic scale. betaDisplacementLog[comp] = beta_displacement_log_plus_2048[comp] – 211 NOTE – the reference rate control parameter beta (β), mentioned here, is the parameter used in the model training to control the ratio between the bitrate and distortion. “The model” here means the model associated with the index of the used model (model_id). When syntax element beta_displacement_log_plus_2048[1] is not present (independent_beta_uv is equal to 0), betaDisplacementLog[1] = betaDisplacementLog[0]. [0054] More details of the embodiments of the present disclosure will be described below which are related to neural network-based visual data coding. As used herein, the term 69 F1250625PCT “visual data” may refer to an image, a video, a picture in a video, or any other visual data suitable to be coded. To solve the above problems and some other problems not mentioned, visual data processing solutions as described below are disclosed. The embodiments of the present disclosure should be considered as examples to explain the general concepts and should not be interpreted in a narrow way. Furthermore, these embodiments can be applied individually or combined in any manner. [0055] Fig. 28 illustrates a flowchart of a method 2800 for visual data processing in accordance with some embodiments of the present disclosure. At 2802, a conversion between the visual data and a codestream of the visual data is performed with a neural network (NN)-based model. For example, a codestream may comprise a sequence of bits. In addition, the codestream may further comprise associated codes which are used as markers. As used herein, the codestream may also be referred to as a bitstream. [0056] In some embodiments, the conversion may include encoding the visual data into the codestream. Additionally or alternatively, the conversion may include decoding the visual data from the codestream. By way of example rather than limitation, the decoding model shown in Fig. 6 may be employed for decoding the visual data from the bitstream. [0057] As used herein, an NN-based model may be a model based on neural network technologies. For example, an NN-based model may specify sequence of neural network modules (also called architecture) and model parameters. The neural network module may comprise a set of neural network layers. Each neural network layer specifies a tensor operation which receives and outputs tensor, and each layer has trainable parameters. It should be understood that the possible implementations of the NN-based model described here are merely illustrative and therefore should not be construed as limiting the present disclosure in any way. [0058] Moreover, the codestream comprises at least one indication associated with a usage of a variable rate coding process in the conversion. In one example embodiment, the at least one indication may comprise a first indication indicating whether the variable rate coding process is enabled or disabled. In this case, it is possible to turn on or turn off the variable rate functionality. Thereby, it is possible to skip operations only necessary for performing variable rate coding process in a case where the variable rate coding process is disabled, and thus the coding efficiency can be improved. It should be noted that the at least one indication may also comprise any other suitable indication(s), a part of which will be described in detail below. [0059] In view of the above, at least one indication associated with a usage of a variable rate coding process in the conversion is signaled in the bitstream. Compared with the 70 F1250625PCT conventional solution, the proposed method can better support the application of variable rate coding process. Thereby, the coding flexibility can be improved. [0060] In some embodiments, the at least one indication may comprise a second indication indicating whether a first parameter or a second parameter is used in the variable rate coding process. The first parameter is predetermined, and the second parameter is obtained based on the codestream. For example, the first parameter and/or the second parameter may be used to adjust a residual parameter or a standard deviation parameter (a.k.a., sigma parameter) for the visual data. By way of example rather than limitation, the residual parameter may comprise a residual tensor, and the standard deviation parameter may comprise a standard deviation logarithm tensor. [0061] In some embodiments, the first parameter may be determined from a plurality of predetermined parameters based on an index. In one example embodiment, the index may comprise a model index. Additionally or alternatively, the index may be indicated in the codestream. [0062] In some embodiments, the second parameter may be determined based on a sum of the first parameter and a value indicated in the codestream. By way of example rather than limitation, the second parameter may be determined as follows: ^^^^^^^^, ^^, ^^^ ൌ ^^^^^^^^^^^^^^^^^ௗ^^^^^^^^^^^^^^^ ^ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ where ^^^^^^^^ represents a logarithmic domain gain tensor, ^^ represents a channel index, ^^ and ^^ represent two spatial indexes, ^^^^^^^^^^^ௗ represents a model index, ^^^^^^^^ represents a color component index, ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ represents a logarithmic domain displacement between the rate control parameter beta selected by encoder for a color component and the reference rate control parameter beta associated with the index of the used model (^^^^^^^^^^^ௗ). In the above equation, ^^^^^^^^, ^^, ^^^ may correspond to the second parameter, ^^^^^^^^^^^^^^^^^ௗ^^^^^^^^^^^^^^^ may correspond to the first parameter, and ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ may be indicated in the codestream. [0063] In some embodiments, in accordance with a determination that the second indication is equal to a first value (such as 0, FALSE, or the like), the first parameter may be used in the variable rate coding process. In accordance with a determination that the second indication is equal to a second value (such as 1, TRUE, or the like), the second parameter may be used in the variable rate coding process. [0064] In some embodiments, the at least one indication may comprise a third indication indicating whether one or more indications associated with the variable rate coding process are comprised in the codestream. For example, the one or more indications comprise at least one of the following: an indication indicating whether a rate control parameter for a primary 71 F1250625PCT component and a rate control parameter for a secondary component are different, an indication indicating a displacement between a rate control parameter selected by an encoder and a reference rate control parameter associated with an index of the used model, and/or the like. In some embodiments, the third indication may be represented as beta_displacement_present_flag, beta_displacement_inferred_flag, variable_rate_coding_enabled flag, or any other suitable string. The scope of the present disclosure is not limited in this respect. [0065] In some embodiments, the at least one indication may comprise a fourth indication indicating whether a value of a third parameter associated with the variable rate coding process is inferred to be a predetermined value. For example, the predetermined value may be 0 or 2048. In one example embodiment, the third parameter may be a displacement between a rate control parameter selected by an encoder and a reference rate control parameter associated with an index of the used model. Alternatively, the third parameter may be the displacement plus 2048. [0066] In some embodiments, the at least one indication may be comprised in a header syntax structure in the codestream. In one example embodiment, the header syntax structure may be a picture header. In the picture header, a model header immediately follows the at least one indication. In another example embodiment, the header syntax structure may be a model header. The at least one indication may be at the initial position within the model header. [0067] In view of the above, the solutions in accordance with some embodiments of the present disclosure can advantageously improve coding efficiency and coding flexibility. [0068] According to further embodiments of the present disclosure, a non-transitory computer-readable recording medium is provided. The non-transitory computer-readable recording medium stores a codestream of visual data which is generated by a method performed by an apparatus for visual data processing. The method comprises: performing a conversion from the visual data to the codestream with a neural network (NN)-based model, wherein the codestream comprises at least one indication associated with a usage of a variable rate coding process in the conversion. [0069] According to still further embodiments of the present disclosure, a method for storing codestream of visual data is provided. The method comprises: performing a conversion from the visual data to the codestream with a neural network (NN)-based model; and storing the codestream in a non-transitory computer-readable recording medium, wherein the codestream comprises at least one indication associated with a usage of a variable rate coding process in the conversion. 72 F1250625PCT [0070] Implementations of the present disclosure can be described in view of the following clauses, the features of which can be combined in any reasonable manner. [0071] Clause 1. A method for visual data processing, comprising: performing a conversion between visual data and a codestream of the visual data with a neural network (NN)-based model, wherein the codestream comprises at least one indication associated with a usage of a variable rate coding process in the conversion. [0072] Clause 2. The method of clause 1, wherein the at least one indication comprises a first indication indicating whether the variable rate coding process is enabled or disabled. [0073] Clause 3. The method of any of clauses 1-2, wherein the at least one indication comprises a second indication indicating whether a first parameter or a second parameter is used in the variable rate coding process, the first parameter being predetermined and the second parameter being obtained based on the codestream. [0074] Clause 4. The method of clause 3, wherein the first parameter or the second parameter is used to adjust a residual parameter or a standard deviation parameter for the visual data. [0075] Clause 5. The method of any of clauses 3-4, wherein the first parameter is determined from a plurality of predetermined parameters based on an index. [0076] Clause 6. The method of clause 5, wherein the index comprises a model index, or the index is indicated in the codestream. [0077] Clause 7. The method of any of clauses 3-6, wherein in accordance with a determination that the second indication is equal to a first value, the first parameter is used in the variable rate coding process, or in accordance with a determination that the second indication is equal to a second value, the second parameter is used in the variable rate coding process. [0078] Clause 8. The method of any of clauses 3-7, wherein the second parameter is determined based on a sum of the first parameter and a value indicated in the codestream. [0079] Clause 9. The method of any of clauses 1-8, wherein the at least one indication comprises a third indication indicating whether one or more indications associated with the variable rate coding process are comprised in the codestream. [0080] Clause 10. The method of clause 9, wherein the one or more indications comprise at least one of the following: an indication indicating whether a rate control parameter for a primary component and a rate control parameter for a secondary component are different, or an indication indicating a displacement between a rate control parameter selected by an encoder and a reference rate control parameter associated with an index of the used model. [0081] Clause 11. The method of any of clauses 9-10, wherein the third indication is 73 F1250625PCT represented as beta_displacement_present_flag, beta_displacement_inferred_flag, or variable_rate_coding_enabled flag. [0082] Clause 12. The method of any of clauses 1-11,wherein the at least one indication comprises a fourth indication indicating whether a value of a third parameter associated with the variable rate coding process is inferred to be a predetermined value. [0083] Clause 13. The method of clause 12, wherein the predetermined value is 0 or 2048. [0084] Clause 14. The method of any of clauses 12-13, wherein the third parameter is a displacement between a rate control parameter selected by an encoder and a reference rate control parameter associated with an index of the used model, or the third parameter is the displacement plus 2048. [0085] Clause 15. The method of any of clauses 1-14, wherein the at least one indication is comprised in a header syntax structure in the codestream. [0086] Clause 16. The method of clause 15, wherein the header syntax structure is a picture header. [0087] Clause 17. The method of clause 16, wherein in the picture header, a model header immediately follows the at least one indication. [0088] Clause 18. The method of clause 15, wherein the header syntax structure is a model header. [0089] Clause 19. The method of clause 18, wherein the at least one indication is at the initial position within the model header. [0090] Clause 20. The method of any of clauses 1-19, wherein the visual data is at least a part of a picture of a video or an image. [0091] Clause 21. The method of any of clauses 1-20, wherein the conversion includes encoding the visual data into the codestream. [0092] Clause 22. The method of any of clauses 1-20, wherein the conversion includes decoding the visual data from the codestream. [0093] Clause 23. An apparatus for visual data processing comprising a processor and a non-transitory memory with instructions thereon, wherein the instructions upon execution by the processor, cause the processor to perform a method in accordance with any of clauses 1-22. [0094] Clause 24. A non-transitory computer-readable storage medium storing instructions that cause a processor to perform a method in accordance with any of clauses 1-22. [0095] Clause 25. A non-transitory computer-readable recording medium storing a codestream of visual data which is generated by a method performed by an apparatus for visual data processing, wherein the method comprises: performing a conversion from the 74 F1250625PCT visual data to the codestream with a neural network (NN)-based model, wherein the codestream comprises at least one indication associated with a usage of a variable rate coding process in the conversion. [0096] Clause 26. A method for storing a codestream of visual data, comprising: performing a conversion from the visual data to the codestream with a neural network (NN)- based model; and storing the codestream in a non-transitory computer-readable recording medium, wherein the codestream comprises at least one indication associated with a usage of a variable rate coding process in the conversion. Example Device [0097] Fig. 29 illustrates a block diagram of a computing device 2900 in which various embodiments of the present disclosure can be implemented. The computing device 2900 may be implemented as or included in the source device 110 (or the visual data encoder 114) or the destination device 120 (or the visual data decoder 124). [0098] It would be appreciated that the computing device 2900 shown in Fig. 29 is merely for purpose of illustration, without suggesting any limitation to the functions and scopes of the embodiments of the present disclosure in any manner. [0099] As shown in Fig. 29, the computing device 2900 includes a general-purpose computing device 2900. The computing device 2900 may at least comprise one or more processors or processing units 2910, a memory 2920, a storage unit 2930, one or more communication units 2940, one or more input devices 2950, and one or more output devices 2960. [0100] In some embodiments, the computing device 2900 may be implemented as any user terminal or server terminal having the computing capability. The server terminal may be a server, a large-scale computing device or the like that is provided by a service provider. The user terminal may for example be any type of mobile terminal, fixed terminal, or portable terminal, including a mobile phone, station, unit, device, multimedia computer, multimedia tablet, Internet node, communicator, desktop computer, laptop computer, notebook computer, netbook computer, tablet computer, personal communication system (PCS) device, personal navigation device, personal digital assistant (PDA), audio/video player, digital camera/video camera, positioning device, television receiver, radio broadcast receiver, E-book device, gaming device, or any combination thereof, including the accessories and peripherals of these devices, or any combination thereof. It would be contemplated that the computing device 2900 can support any type of interface to a user (such as “wearable” circuitry and the like). [0101] The processing unit 2910 may be a physical or virtual processor and can implement 75 F1250625PCT various processes based on programs stored in the memory 2920. In a multi-processor system, multiple processing units execute computer executable instructions in parallel so as to improve the parallel processing capability of the computing device 2900. The processing unit 2910 may also be referred to as a central processing unit (CPU), a microprocessor, a controller or a microcontroller. [0102] The computing device 2900 typically includes various computer storage medium. Such medium can be any medium accessible by the computing device 2900, including, but not limited to, volatile and non-volatile medium, or detachable and non-detachable medium. The memory 2920 can be a volatile memory (for example, a register, cache, Random Access Memory (RAM)), a non-volatile memory (such as a Read-Only Memory (ROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), or a flash memory), or any combination thereof. The storage unit 2930 may be any detachable or non-detachable medium and may include a machine-readable medium such as a memory, flash memory drive, magnetic disk or another other media, which can be used for storing information and/or visual data and can be accessed in the computing device 2900. [0103] The computing device 2900 may further include additional detachable/non- detachable, volatile/non-volatile memory medium. Although not shown in Fig. 29, it is possible to provide a magnetic disk drive for reading from and/or writing into a detachable and non-volatile magnetic disk and an optical disk drive for reading from and/or writing into a detachable non-volatile optical disk. In such cases, each drive may be connected to a bus (not shown) via one or more visual data medium interfaces. [0104] The communication unit 2940 communicates with a further computing device via the communication medium. In addition, the functions of the components in the computing device 2900 can be implemented by a single computing cluster or multiple computing machines that can communicate via communication connections. Therefore, the computing device 2900 can operate in a networked environment using a logical connection with one or more other servers, networked personal computers (PCs) or further general network nodes. [0105] The input device 2950 may be one or more of a variety of input devices, such as a mouse, keyboard, tracking ball, voice-input device, and the like. The output device 2960 may be one or more of a variety of output devices, such as a display, loudspeaker, printer, and the like. By means of the communication unit 2940, the computing device 2900 can further communicate with one or more external devices (not shown) such as the storage devices and display device, with one or more devices enabling the user to interact with the computing device 2900, or any devices (such as a network card, a modem and the like) enabling the computing device 2900 to communicate with one or more other computing 76 F1250625PCT devices, if required. Such communication can be performed via input/output (I/O) interfaces (not shown). [0106] In some embodiments, instead of being integrated in a single device, some or all components of the computing device 2900 may also be arranged in cloud computing architecture. In the cloud computing architecture, the components may be provided remotely and work together to implement the functionalities described in the present disclosure. In some embodiments, cloud computing provides computing, software, visual data access and storage service, which will not require end users to be aware of the physical locations or configurations of the systems or hardware providing these services. In various embodiments, the cloud computing provides the services via a wide area network (such as Internet) using suitable protocols. For example, a cloud computing provider provides applications over the wide area network, which can be accessed through a web browser or any other computing components. The software or components of the cloud computing architecture and corresponding visual data may be stored on a server at a remote position. The computing resources in the cloud computing environment may be merged or distributed at locations in a remote visual data center. Cloud computing infrastructures may provide the services through a shared visual data center, though they behave as a single access point for the users. Therefore, the cloud computing architectures may be used to provide the components and functionalities described herein from a service provider at a remote location. Alternatively, they may be provided from a conventional server or installed directly or otherwise on a client device. [0107] The computing device 2900 may be used to implement visual data encoding/decoding in embodiments of the present disclosure. The memory 2920 may include one or more visual data coding modules 2925 having one or more program instructions. These modules are accessible and executable by the processing unit 2910 to perform the functionalities of the various embodiments described herein. [0108] In the example embodiments of performing visual data encoding, the input device 2950 may receive visual data as an input 2970 to be encoded. The visual data may be processed, for example, by the visual data coding module 2925, to generate an encoded bitstream. The encoded bitstream may be provided via the output device 2960 as an output 2980. [0109] In the example embodiments of performing visual data decoding, the input device 2950 may receive an encoded bitstream as the input 2970. The encoded bitstream may be processed, for example, by the visual data coding module 2925, to generate decoded visual data. The decoded visual data may be provided via the output device 2960 as the output 2980. 77 F1250625PCT [0110] While this disclosure has been particularly shown and described with references to example embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the present application as defined by the appended claims. Such variations are intended to be covered by the scope of this present application. As such, the foregoing description of embodiments of the present application is not intended to be limiting. 78 F1250625PCT

Claims

I/We Claim: 1. A method for visual data processing, comprising: performing a conversion between visual data and a codestream of the visual data with a neural network (NN)-based model, wherein the codestream comprises at least one indication associated with a usage of a variable rate coding process in the conversion. 2. The method of claim 1, wherein the at least one indication comprises a first indication indicating whether the variable rate coding process is enabled or disabled. 3. The method of any of claims 1-2, wherein the at least one indication comprises a second indication indicating whether a first parameter or a second parameter is used in the variable rate coding process, the first parameter being predetermined and the second parameter being obtained based on the codestream. 4. The method of claim 3, wherein the first parameter or the second parameter is used to adjust a residual parameter or a standard deviation parameter for the visual data. 5. The method of any of claims 3-4, wherein the first parameter is determined from a plurality of predetermined parameters based on an index. 6. The method of claim 5, wherein the index comprises a model index, or the index is indicated in the codestream. 7. The method of any of claims 3-6, wherein in accordance with a determination that the second indication is equal to a first value, the first parameter is used in the variable rate coding process, or in accordance with a determination that the second indication is equal to a second value, the second parameter is used in the variable rate coding process. 8. The method of any of claims 3-7, wherein the second parameter is determined based on a sum of the first parameter and a value indicated in the codestream. 9. The method of any of claims 1-8, wherein the at least one indication comprises a third indication indicating whether one or more indications associated with the variable rate 79 F1250625PCT coding process are comprised in the codestream. 10. The method of claim 9, wherein the one or more indications comprise at least one of the following: an indication indicating whether a rate control parameter for a primary component and a rate control parameter for a secondary component are different, or an indication indicating a displacement between a rate control parameter selected by an encoder and a reference rate control parameter associated with an index of the used model. 11. The method of any of claims 9-10, wherein the third indication is represented as beta_displacement_present_flag, beta_displacement_inferred_flag, or variable_rate_coding_enabled flag. 12. The method of any of claims 1-11,wherein the at least one indication comprises a fourth indication indicating whether a value of a third parameter associated with the variable rate coding process is inferred to be a predetermined value. 13. The method of claim 12, wherein the predetermined value is 0 or 2048. 14. The method of any of claims 12-13, wherein the third parameter is a displacement between a rate control parameter selected by an encoder and a reference rate control parameter associated with an index of the used model, or the third parameter is the displacement plus 2048. 15. The method of any of claims 1-14, wherein the at least one indication is comprised in a header syntax structure in the codestream. 16. The method of claim 15, wherein the header syntax structure is a picture header. 17. The method of claim 16, wherein in the picture header, a model header immediately follows the at least one indication. 18. The method of claim 15, wherein the header syntax structure is a model header. 19. The method of claim 18, wherein the at least one indication is at the initial position 80 F1250625PCT within the model header. 20. The method of any of claims 1-19, wherein the visual data is at least a part of a picture of a video or an image. 21. The method of any of claims 1-20, wherein the conversion includes encoding the visual data into the codestream. 22. The method of any of claims 1-20, wherein the conversion includes decoding the visual data from the codestream. 23. An apparatus for visual data processing comprising a processor and a non- transitory memory with instructions thereon, wherein the instructions upon execution by the processor, cause the processor to perform a method in accordance with any of claims 1-22. 24. A non-transitory computer-readable storage medium storing instructions that cause a processor to perform a method in accordance with any of claims 1-22. 25. A non-transitory computer-readable recording medium storing a codestream of visual data which is generated by a method performed by an apparatus for visual data processing, wherein the method comprises: performing a conversion from the visual data to the codestream with a neural network (NN)-based model, wherein the codestream comprises at least one indication associated with a usage of a variable rate coding process in the conversion. 26. A method for storing a codestream of visual data, comprising: performing a conversion from the visual data to the codestream with a neural network (NN)-based model; and storing the codestream in a non-transitory computer-readable recording medium, wherein the codestream comprises at least one indication associated with a usage of a variable rate coding process in the conversion. 81 F1250625PCT
PCT/US2025/023092 2024-04-04 2025-04-04 Method, apparatus, and medium for visual data processing Pending WO2025212980A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US202463574644P 2024-04-04 2024-04-04
US63/574,644 2024-04-04

Publications (1)

Publication Number Publication Date
WO2025212980A1 true WO2025212980A1 (en) 2025-10-09

Family

ID=97268016

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2025/023092 Pending WO2025212980A1 (en) 2024-04-04 2025-04-04 Method, apparatus, and medium for visual data processing

Country Status (1)

Country Link
WO (1) WO2025212980A1 (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180240221A1 (en) * 2017-02-17 2018-08-23 Cogisen S.R.L. Method for image processing and video compression
US10282821B1 (en) * 2015-08-27 2019-05-07 Descartes Labs, Inc. Observational data processing and analysis
US20190244394A1 (en) * 2018-02-08 2019-08-08 Uber Technologies, Inc. Leveraging jpeg discrete cosine transform coefficients in neural networks
US20200092556A1 (en) * 2018-09-18 2020-03-19 Google Llc Efficient Use of Quantization Parameters in Machine-Learning Models for Video Coding
US20200304147A1 (en) * 2019-03-18 2020-09-24 Samsung Electronics Co., Ltd. Method and apparatus for variable rate compression with a conditional autoencoder

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10282821B1 (en) * 2015-08-27 2019-05-07 Descartes Labs, Inc. Observational data processing and analysis
US20180240221A1 (en) * 2017-02-17 2018-08-23 Cogisen S.R.L. Method for image processing and video compression
US20190244394A1 (en) * 2018-02-08 2019-08-08 Uber Technologies, Inc. Leveraging jpeg discrete cosine transform coefficients in neural networks
US20200092556A1 (en) * 2018-09-18 2020-03-19 Google Llc Efficient Use of Quantization Parameters in Machine-Learning Models for Video Coding
US20200304147A1 (en) * 2019-03-18 2020-09-24 Samsung Electronics Co., Ltd. Method and apparatus for variable rate compression with a conditional autoencoder

Similar Documents

Publication Publication Date Title
US20260019577A1 (en) Method, apparatus, and medium for visual data processing
WO2025072500A1 (en) Method, apparatus, and medium for visual data processing
WO2024140849A9 (en) Method, apparatus, and medium for visual data processing
US20250373827A1 (en) Method, apparatus, and medium for visual data processing
US20250379990A1 (en) Method, apparatus, and medium for visual data processing
WO2025198937A1 (en) Method, apparatus, and medium for visual data processing
WO2024149392A1 (en) Method, apparatus, and medium for visual data processing
WO2025149063A1 (en) Method, apparatus, and medium for visual data processing
WO2025157163A1 (en) Method, apparatus, and medium for visual data processing
WO2025212980A1 (en) Method, apparatus, and medium for visual data processing
WO2025200931A1 (en) Method, apparatus, and medium for visual data processing
WO2025131046A1 (en) Method, apparatus, and medium for visual data processing
WO2025146073A1 (en) Method, apparatus, and medium for visual data processing
WO2024193710A1 (en) Method, apparatus, and medium for visual data processing
WO2025077746A1 (en) Method, apparatus, and medium for visual data processing
WO2025077742A1 (en) Method, apparatus, and medium for visual data processing
WO2025077744A1 (en) Method, apparatus, and medium for visual data processing
WO2025082522A1 (en) Method, apparatus, and medium for visual data processing
WO2025082523A1 (en) Method, apparatus, and medium for visual data processing
WO2025044947A1 (en) Method, apparatus, and medium for visual data processing
WO2025153016A1 (en) Method, apparatus, and medium for visual data processing
WO2024083202A1 (en) Method, apparatus, and medium for visual data processing
WO2025087230A1 (en) Method, apparatus, and medium for visual data processing
WO2025002424A1 (en) Method, apparatus, and medium for visual data processing
WO2024193708A1 (en) Method, apparatus, and medium for visual data processing

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 25783363

Country of ref document: EP

Kind code of ref document: A1