CN115243043A - Depth image coding and decoding method and system with controllable entropy decoding complexity - Google Patents
Depth image coding and decoding method and system with controllable entropy decoding complexity Download PDFInfo
- Publication number
- CN115243043A CN115243043A CN202210883754.0A CN202210883754A CN115243043A CN 115243043 A CN115243043 A CN 115243043A CN 202210883754 A CN202210883754 A CN 202210883754A CN 115243043 A CN115243043 A CN 115243043A
- Authority
- CN
- China
- Prior art keywords
- autoregressive
- entropy
- decoding
- probability
- model
- 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
Links
Images
Classifications
-
- 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
- H04N19/13—Adaptive entropy coding, e.g. adaptive variable length coding [AVLC] or context adaptive binary arithmetic coding [CABAC]
-
- 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
- H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
- H04N19/146—Data rate or code amount at the encoder output
- H04N19/149—Data rate or code amount at the encoder output by estimating the code amount by means of a model, e.g. mathematical model or statistical model
-
- 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
- H04N19/44—Decoders specially adapted therefor, e.g. video decoders which are asymmetric with respect to the encoder
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Algebra (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention discloses a depth image coding and decoding method and system with controllable entropy decoding complexity, which can support the requirements of different decoding complexity in the same model by introducing an autoregressive reference relationship, save storage consumption and improve the practicability of a mainstream depth image coding and decoding frame; after the autoregressive reference relationship is introduced, the extra code stream occupied by the index for transmitting the autoregressive reference relationship is very small, and the autoregressive reference relationship has good compression performance while supporting multiple complexities.
Description
Technical Field
The invention relates to the technical field of image compression coding, in particular to a depth image coding and decoding method and system with controllable entropy decoding complexity.
Background
Image compression is an important technique for reducing the consumption of image storage space and transmission bandwidth, and the time complexity is one of the determining factors in selecting to deploy a lossy image codec in practical applications.
The conventional image codecs widely used at present, such as HEVC/h.265intra and VVC/h.266intra, pursue better rate-distortion performance by selecting the best coding mode from a large number of candidate coding modes, but bring about serious time consumption. Although the depth image coding methods combined with deep learning in recent years achieve impressive rate-distortion performance, even better than the most advanced traditional image codec, their decoding time is often much longer.
Previous work has shown that a complex but efficient entropy model is crucial to improving rate-distortion performance, but at the expense of unacceptable decoding time. The commonly used spatial autoregressive context entropy model can bring about a rate savings of about 11%, but due to the need for serial decoding, even with fine engineering optimization, it still takes more than 7 seconds to decode the image at standard 1080p resolution. For practical use, some recent studies have explored some effective alternatives to such a spatial autoregressive model, but all pursue the best rate-distortion performance at the inherent decoding complexity, and do not consider the requirement of supporting different complexities in one model.
Disclosure of Invention
The invention aims to provide a depth image coding and decoding method and system with controllable entropy decoding complexity, which use a multi-complexity entropy decoding scheme with high expandability to enable a decoder to support any predefined good complexity requirement.
The purpose of the invention is realized by the following technical scheme:
a depth image coding and decoding method with controllable entropy decoding complexity is characterized by comprising the following steps:
inputting an original image x, converting the original image x into an original characteristic y through an encoder, and inputting the original characteristic y into two branches; the first branch is a super-first-check branch, and the super-first-check branch outputs a super-first-check characteristic
In the second branch, the original characteristic y is quantized into discrete representation
And inputting the data into an entropy model and an autoregressive model respectively, wherein the autoregressive model is based on an autoregressive reference relation and a super-first-order characteristic specified by a user
Performing autoregressive probability modeling to predict discrete characterization
Probability distribution of
Incorporating probability distributions from the entropy model
Characterizing said dispersion
Entropy coding is carried out, and entropy decoding is carried out by combining the corresponding autoregressive reference relation; inverse transformation of entropy-decoded representations into reconstructed images by a decoder
Wherein different autoregressive reference relationships correspond to different complexities.
An entropy decoding complexity controllable depth image coding and decoding system, comprising: the device comprises an encoder, a first branch circuit, a second branch circuit and a decoder; wherein:
an encoder for transforming an input original image x into an original feature y;
the first branch is a super-first-check branch, and outputs a super-first-check characteristic
In the second branch, the primitive character is firstQuantification of token y into discrete tokens
Respectively inputting the data into an entropy model and an autoregressive model, wherein the autoregressive model is based on the autoregressive reference relation and the super-prior-experience characteristic specified by a user
Performing autoregressive probability modeling to predict discrete characterization
Probability distribution of
Incorporating probability distributions from the entropy model
Characterizing said dispersion
Entropy encoding and decoding are carried out; wherein different autoregressive reference relationships correspond to different complexities;
decoder for inverse transformation of entropy decoded representations into reconstructed images
A processing device, comprising: one or more processors; a memory for storing one or more programs;
wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the aforementioned method.
A readable storage medium, storing a computer program which, when executed by a processor, implements the aforementioned method.
According to the technical scheme provided by the invention, the requirements of different decoding complexity can be supported in the same model by introducing the autoregressive reference relation, so that the storage consumption is saved, and the practicability of a mainstream depth image coding and decoding frame is improved; after the autoregressive reference relationship is introduced, the extra code stream occupied by the index for transmitting the autoregressive reference relationship is very small, and the autoregressive reference relationship has good compression performance while supporting multiple complexities.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on the drawings without creative efforts.
Fig. 1 is a schematic diagram of a depth image coding and decoding method with controllable entropy decoding complexity according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a checkerboard context model provided by an embodiment of the present invention;
FIG. 3 is a diagram of a joint checkerboard context and inter-row autoregressive model according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a channel autoregressive model according to an embodiment of the present invention;
FIG. 5 is a diagram illustrating a spatial autoregressive context model according to an embodiment of the present invention;
fig. 6 is a schematic diagram of a depth image coding and decoding system with controllable entropy decoding complexity according to an embodiment of the present invention;
FIG. 7 is a schematic diagram of a processing apparatus according to an embodiment of the present invention;
Detailed Description
The technical solutions in the embodiments of the present invention are clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present invention without making any creative effort, shall fall within the protection scope of the present invention.
The terms that may be used herein are first described as follows:
the term "and/or" means that either or both can be achieved, for example, X and/or Y means that both cases include "X" or "Y" as well as three cases including "X and Y".
The terms "comprising," "including," "containing," "having," or other similar terms of meaning should be construed as non-exclusive inclusions. For example: including a feature (e.g., material, component, ingredient, carrier, formulation, material, dimension, part, component, mechanism, device, process, procedure, method, reaction condition, processing condition, parameter, algorithm, signal, data, product, or article of manufacture), is to be construed as including not only the particular feature explicitly listed but also other features not explicitly listed as such which are known in the art.
The depth image coding and decoding method and system with controllable entropy decoding complexity provided by the invention are described in detail below. Details which are not described in detail in the embodiments of the invention belong to the prior art which is known to a person skilled in the art. The examples of the present invention, in which specific conditions are not specified, were carried out according to the conventional conditions in the art or conditions suggested by the manufacturer.
Example one
The embodiment of the invention provides a depth image coding and decoding method with controllable entropy decoding complexity, an integral framework of the method is shown in figure 1, and the integral process comprises the following steps: inputting an original image x, converting the original image x into an original characteristic y through an encoder, and inputting the original characteristic y into two branches; the first branch is the super-first branch (right part of figure 1)
Branch of) from the super-first branch, outputting the super-first characteristics from the super-first branch
In the second branch, the original characteristic y is quantized into discrete representation
Respectively inputting the data into an entropy model and an autoregressive model, wherein the autoregressive model is based on the autoregressive reference relation and the super-prior-experience characteristic specified by a user
Performing autoregressive probability modeling to predict discrete characterization
Probability distribution of
Incorporating probability distributions by the entropy model
Characterizing said dispersion
Entropy coding is carried out, and entropy decoding is carried out by combining the corresponding autoregressive reference relation; inverse transformation of entropy-decoded representations into reconstructed images by a decoder
Wherein different autoregressive reference relationships correspond to different complexities.
In the above flows, the processing flow related to the super-first-check branch can refer to the conventional technology, and is not described in detail in the present invention; entropy decoding and entropy encoding have a variety of approaches, and FIG. 1 provides an example of arithmetic encoding and arithmetic decoding; moreover, lossless entropy coding and lossless entropy decoding are used in both branches, and thus, the object of entropy coding (discrete representation)
And
) I.e. a representation obtained by entropy decoding.
For ease of understanding, the foregoing aspects of the invention are described in detail below.
1. And (5) encoding and decoding processes after an autoregressive reference relation is introduced.
The transformation and inverse transformation involved in the encoding and decoding process of the invention and the processing of the super-prior branch (the whole part of the branch can be called as a super-prior model) are similar to the traditional depth image encoding and decoding scheme. However, the conventional depth image coding and decoding scheme does not consider the autoregressive reference relationship, and in the second branch, the entropy model is used for estimating
Using only the decoded token set in the current autoregressive model
And the super-check characteristics obtained by the inverse super-check transform
Due to super-prior sub-branch
And the occupied time complexity is low due to parallel processing in the coding and decoding processes. But the autoregressive branch in the second branch
Because of the constraint of autoregressive property, serial decoding is required at a decoding end, most of time of entropy decoding is occupied, and the complexity T of entropy decoding is e Positively correlated with the autoregressive degree a:
T e ＝O(a)。
where O () is a common temporal complexity symbol that describes the magnitude of the complexity.
In order to support different requirements of entropy decoding complexity by a depth image coding and decoding framework, the entropy decoding complexity is controllable. In the embodiment of the invention, a series of autoregressive reference relationship sets R, R = { R } are predefined 1 ,R 2 ,…,R M Defining reference index i =1,2, …, M, M is the number of autoregressive reference relations, autoregressiveEach item R in the set of reference relationships i All the parameters are autoregressive reference relations, different autoregressive reference relations correspond to different autoregressive times a, and the autoregressive times determine the complexity of entropy decoding. During entropy coding, a user can select a reference relation R corresponding to the complexity according to the required target decoding complexity T i And performing autoregressive probability modeling, and compressing a reference index i corresponding to an autoregressive reference relation into a code stream by the entropy model for transmission. When entropy decoding, searching corresponding R according to the solved reference index i i The autoregressive entropy decoding can be performed. In the code stream, a bit transmission index i with a set length is reserved, and the bits occupied by the reference index are related to the number M of the autoregressive reference relations. In the embodiment of the present invention, the number of the autoregressive reference relationships is M, and the autoregressive reference relationship set R may vary according to actual use requirements. The number of bits consumed to transmit an index is log 2 M, for example: when M =2, the autoregressive reference relation set R only comprises two modes, and a 1-bit transmission index is reserved; when M =128, the set of autoregressive reference relations R contains 128 patterns, we will reserve a 7-bit transmission index. The total code stream comprises: code stream of reference index i, characterization
Code stream of shape information, code representation
Code stream of itself, characterization
Code stream of shape information, code representation
The code stream itself is not described in detail in consideration that the latter four code streams are also included in the existing deep coding and decoding model.
2. And (4) introducing an autoregressive reference relationship to obtain a probability distribution calculation scheme.
In the embodiment of the invention, the autoregressive reference relation R i Is a and discrete characterization
The value of each point in the autoregressive reference relation is an integer which is greater than or equal to zero and represents a discrete characterization
The representation points of the stomach corresponding to the middle position have precedence relationship in the autoregressive decoding process. For example, autoregressive reference relationship R i When all the values are 0, all the characteristics can be entropy decoded without an autoregressive process, and in the embodiment of the invention, decoding only using a super-prior model is adopted; and autoregressive reference relation R i Discrete characterization of corresponding location of value 1 when one portion is 0 and one portion is 1
The upper token point will depend on the discrete token of the corresponding position with value 0 during entropy decoding
And the autoregressive degree a of the upper characterization point is 1. Based on this, the discrete characterization can be performed
The entropy decoding is divided into K groups in the order of entropy decoding, represented as:
C k ＝{v|R i (v)＝k}，k＝0，…，K-1
wherein, the first and the second end of the pipe are connected with each other,
representing discrete tokens
V denotes the coordinates of the token vector, R i Representing an autoregressive reference relationship, R i (v) = k denotes autoregressive reference relation R i The representative vector coordinates of all points with the value k in the list, C k Representing an autoregressive reference relationship R i A set of characterization vector coordinates of all points with the value of k, and discrete characterization
The kth group of eigenvectors in (1)
Is set C k And (6) correspondingly characterizing.
Based on the discrete characterization
Form of (1), probability distribution thereof
The predicting step mainly comprises:
(1) Estimation module g by probability in the super-a branch p The probability parameter Φ is estimated, expressed as:
Φ＝{Φ 0 ，…，Φ K-1 }
wherein phi k Representing the k-th set of eigenvectors
Exemplary:
mean value μ k And variance σ k All probability parameters are probability parameters, and certainly, gaussian distribution is taken as an example here, if other distributions are used, the related probability parameters can be correspondingly adjusted;
represents the decoded representation in the autoregressive model, expressed as:
(2) Performing autoregressive probability modeling on the probability parameters, predicting probability distribution by convolution with unit uniform distribution, and performing characteristic vector calculation on the kth group
The probability distribution is
The calculation process is expressed as:
where P (·) represents autoregressive probabilistic modeling, U (-a, a) represents unit uniform distribution, -a and a represent lower and upper limits of a uniform distribution interval, respectively, and symbol denotes a convolution operation, which is exemplary: u (-A, A) = U (-0.5,0.5).
(3) The probability distribution of all the group characterization vectors is synthesized to obtain discrete characterization
Probability distribution of
In the embodiment of the present invention, the model used for autoregressive probability modeling may be selected by a user according to a situation or experience, for example, a Gaussian distribution model, a Gaussian mixture probability (GMM) model, a Laplace distribution (Laplace distribution) model, or the like may be selected.
3. Example of autoregressive reference relationships.
Examples of five autoregressive references are provided below, corresponding to five entropy decoding complexities from low to high, respectively, and the models mentioned in the following five example headings are the second branch portions described above.
Example one, only the super-prior model is used.
If only the superior model is used, and no prediction mode of the autoregressive model is selected, then the autoregressive reference relation R is used at the moment i ＝0，a＝0。
Example two, a checkerboard context model (Checker-cube context model).
This example presents a checkerboard context model that utilizes discrete representations
The entropy decoding can be completed only by performing an autoregressive process once in the space and cross-channel relation. Corresponding to autoregressive reference relationship R i Is as shown in fig. 2, and at this time a =1.
Example three, a joint checkerboard context and an interline autoregressive model.
On the basis of the context model of checkerboard squares as proposed in example 2, the autoregressive reference relationship R is determined i Of corresponding positions in which the median value is all 0
The above characterization points further increase the dependency relationship. Specifically, an autoregressive reference relation R of a checkerboard context model is extracted first i The corresponding characterization points with the median value of 0 in the singular channel and the even channel are then respectively subjected to the inter-row autoregressive entropy coding, as shown in FIG. 3, at this time, the autoregressive reference relationship R i The middle original value of 0 will be updated by the row number, and the lower row will depend on the upper row when decoding. Assumption of discrete characterization
The total number of rows is W, which increases W-1 times of inter-row autoregressive coding and complexity compared with the original modela＝W。
Example four, channel-wise autoregressive model.
Channel autoregressive model discrete characterization
The channels are divided into S +1 groups, and the decoding of the channels of the S (S =0,1.., S) group depends on the decoding of the channels of the previous S-1 group, but the channels are decoded in parallel, so that the autoregressive reference relation R is obtained at the moment i The values within each set of channels are the same. Corresponding to the autoregressive reference relation R i Is as shown in fig. 4 (i.e. the value of group S is S) and at this time a = S +1.
Example five, a spatial autoregressive context model.
In practical application, the spatial autoregressive context model is an operation mode of predicting a current element by using only left and upper tensor elements of a current spatial position, and serial entropy decoding needs to be performed in the sequence from left to right and from top to bottom during decoding. Therefore, when entropy decoding a discrete representation of H × W size, the spatial autoregressive context model needs to continuously calculate H × W-1 convolutions, and at the moment, the spatial autoregressive context model corresponds to an autoregressive reference relation R i The values of (a) are shown in fig. 5 (i.e., sorted sequentially from left to right, and the next row is sorted sequentially on the basis of the rightmost position of the previous row), and at this time a = H × W-1.
It should be noted that although the models related to examples one, four and five are existing models, the existing work does not introduce the autoregressive reference relationship R i In the above description of the first, fourth and fifth examples, the autoregressive reference relationship R according to the present invention is described i The method and the device are introduced into the existing model, so that the entropy decoding complexity is controllable, namely, the scheme of the invention can be compatible with any existing or future autoregressive entropy model.
The scheme of the embodiment of the invention mainly has the following beneficial effects:
(1) By precisely defining hidden variable characterizations (i.e., discrete characterizations as described above)
) The mathematical expression of the internal reference relation and the utilization of a series of pre-designed autoregressive reference relations can support the requirements of different decoding complexity in the same model, save storage consumption and improve the practicability of the mainstream depth image coding and decoding frame.
(2) The coding and decoding scheme provided by the invention has high expandability, can support any predefined good reference relation except the five reference relation examples provided in the foregoing, occupies a small extra code stream occupied by the transmission reference relation index, and has good compression performance while supporting multiple complexities.
(3) The method has wide applicability and has good function realization on multi-code rate points and multi-data sets.
In conclusion, the coding and decoding scheme provided by the invention can provide better flexibility, practicability and wide applicability in the practical application of the depth image compression framework.
Example two
The present invention further provides a depth image coding and decoding system with controllable entropy decoding complexity, which is implemented mainly based on the method provided by the foregoing embodiment, as shown in fig. 6, the system mainly includes: the device comprises an encoder, a first branch circuit, a second branch circuit and a decoder; wherein:
an encoder for transforming an input original image x into an original feature y;
the first branch is a super-priority branch, and outputs a super-priority characteristic
In the second branch, the original characteristic y is quantized into discrete representation
Respectively inputting the data into an entropy model and an autoregressive model, wherein the autoregressive model is based on the autoregressive reference relation and the super-prior-experience characteristic specified by a user
Performing autoregressive probability modeling to predict discrete characterization
Probability distribution of
Incorporating probability distributions from the entropy model
Characterizing said dispersion
Entropy encoding and decoding are carried out; wherein different autoregressive reference relationships correspond to different complexities;
decoder for inverse transformation of entropy decoded representations into reconstructed images
It will be clear to those skilled in the art that, for convenience and simplicity of description, the foregoing division of the functional modules is merely used as an example, and in practical applications, the above function distribution may be performed by different functional modules according to needs, that is, the internal structure of the system is divided into different functional modules to perform all or part of the above described functions.
EXAMPLE III
The present invention also provides a processing apparatus, as shown in fig. 7, which mainly includes: one or more processors; a memory for storing one or more programs; wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the methods provided by the foregoing embodiments.
Further, the processing device further comprises at least one input device and at least one output device; in the processing device, a processor, a memory, an input device and an output device are connected through a bus.
In the embodiment of the present invention, the specific types of the memory, the input device, and the output device are not limited; for example:
the input device can be a touch screen, an image acquisition device, a physical button or a mouse and the like;
the output device may be a display terminal;
the Memory may be a Random Access Memory (RAM) or a non-volatile Memory (non-volatile Memory), such as a disk Memory.
Example four
The present invention also provides a readable storage medium storing a computer program which, when executed by a processor, implements the method provided by the foregoing embodiments.
The readable storage medium in the embodiment of the present invention may be provided in the foregoing processing device as a computer readable storage medium, for example, as a memory in the processing device. The readable storage medium may be various media that can store program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a magnetic disk, or an optical disk.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (9)
1. A depth image coding and decoding method with controllable entropy decoding complexity is characterized by comprising the following steps:
inputting an original image x, converting the original image x into an original characteristic y through an encoder, and inputting the original characteristic y into two branches; the first branch is a super-pilot branch, and the super-pilot branch outputs a super-pilot characteristic
In the second branch, the original characteristic y is quantized into discrete representation
Respectively inputting the data into an entropy model and an autoregressive model, wherein the autoregressive model is based on the autoregressive reference relation and the super-prior-experience characteristic specified by a user
Performing autoregressive probability modeling to predict discrete characterization
Probability distribution of
Incorporating probability distributions by the entropy model
Characterizing said dispersion
Entropy coding is carried out, and entropy decoding is carried out by combining the corresponding autoregressive reference relation; inverse transformation of entropy-decoded representations into reconstructed images by a decoder
Wherein different autoregressive reference relationships correspond to different complexities.
2. An entropy decoding complexity controllable depth image coding and decoding method according to claim 1, wherein the user-specified autoregressive reference relationship is one of a set of pre-established autoregressive reference relationships R;
the set of autoregressive reference relationships R is represented as: r = { R 1 ，R 2 ，...，R M Define a reference index i =1,2, M being the number of autoregressive reference relations, each term in the set of autoregressive reference relationsR i All are autoregressive reference relations, different autoregressive reference relations correspond to different autoregressive times, and the autoregressive times determine the complexity of entropy decoding.
3. The method as claimed in claim 1 or 2, wherein the entropy model compresses the reference index corresponding to the autoregressive reference relationship into the code stream during entropy encoding, and searches the corresponding autoregressive reference relationship according to the reference index for entropy decoding during entropy decoding.
4. An entropy decoding complexity controlled depth image coding and decoding method according to claim 1 or 2, wherein the autoregressive reference relationship is a relationship with discrete characteristics
The values of all points in the autoregressive reference relation are integers which are more than or equal to zero and represent discrete representations
The sequence relation of the characterization points of the corresponding positions in the autoregressive decoding process.
5. An entropy decoding complexity controlled depth image coding and decoding method according to claim 4, wherein the discrete characterization
The entropy decoding is divided into K groups in the order of entropy decoding, which is expressed as:
C k ＝{v|R i (v)＝k}，k＝0，…，K-1
wherein the content of the first and second substances,
representing discrete tokens
V denotes the coordinates of the token vector, R i Representing an autoregressive reference relationship, R i (v) = k denotes autoregressive reference relation R i The representative vector coordinates of all points with the value k in the list, C k Representing an autoregressive reference relationship R i The set of the characterization vector coordinates of all the points with the value of k, and the discrete characterization
The kth group of eigenvectors in (1)
Is set C k And (6) correspondingly characterizing.
6. An entropy decoding complexity controlled depth image coding and decoding method according to claim 5, characterized in that the discrete characterization is predicted
Probability distribution of
Comprises the following steps:
estimation module g by probability in the super-a branch p The probability parameter Φ is estimated, expressed as:
Φ＝{Φ 0 ，…，Φ K-1 }
wherein phi k Representing the k-th set of eigenvectors
A probability parameter of (d);
represents the decoded representation in the autoregressive model, expressed as:
performing autoregressive probability modeling on the probability parameters, predicting probability distribution by convolution with unit uniform distribution, and performing characteristic vector calculation on the kth group
The probability distribution is
The calculation process is expressed as:
wherein, P (.) represents autoregressive probability modeling, U (-A, A) represents unit uniform distribution, A and A respectively represent a lower limit value and an upper limit value of the unit uniform distribution, and symbol denotes convolution operation;
the probability distribution of all the group characterization vectors is synthesized to obtain discrete characterization
Probability distribution of
7. An entropy decoding complexity-controllable depth image coding and decoding system, which is implemented based on the method of any one of claims 1 to 6, and comprises: the device comprises an encoder, a first branch circuit, a second branch circuit and a decoder; wherein:
an encoder for transforming an input original image x into an original feature y;
the first branch is a super-first-check branch, and outputs a super-first-check characteristic
In the second branch, the original characteristic y is quantized into discrete representation
And inputting the data into an entropy model and an autoregressive model respectively, wherein the autoregressive model is based on an autoregressive reference relation and a super-first-order characteristic specified by a user
Performing autoregressive probability modeling to predict discrete characterization
Probability distribution of
Incorporating probability distributions by the entropy model
For the discrete characterization
Entropy encoding and decoding are carried out; wherein different autoregressive reference relationships correspond to different complexities;
decoder for inverse transformation of entropy decoded representations into reconstructed images
8. A processing device, comprising: one or more processors; a memory for storing one or more programs;
wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of any of claims 1-6.
9. A readable storage medium, storing a computer program, characterized in that the computer program, when being executed by a processor, carries out the method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210883754.0A CN115243043A (en) | 2022-07-26 | 2022-07-26 | Depth image coding and decoding method and system with controllable entropy decoding complexity |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210883754.0A CN115243043A (en) | 2022-07-26 | 2022-07-26 | Depth image coding and decoding method and system with controllable entropy decoding complexity |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115243043A true CN115243043A (en) | 2022-10-25 |
Family
ID=83674518
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210883754.0A Pending CN115243043A (en) | 2022-07-26 | 2022-07-26 | Depth image coding and decoding method and system with controllable entropy decoding complexity |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115243043A (en) |
-
2022
- 2022-07-26 CN CN202210883754.0A patent/CN115243043A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hu et al. | Learning end-to-end lossy image compression: A benchmark | |
| US10623775B1 (en) | End-to-end video and image compression | |
| CN113766228B (en) | Point cloud compression method, encoder, decoder, and storage medium | |
| CN108090521B (en) | Image fusion method and discriminator of generative confrontation network model | |
| CN104641643A (en) | Decomposition of residual data during signal encoding, decoding and reconstruction in a tiered hierarchy | |
| CN111641832A (en) | Encoding method, decoding method, device, electronic device and storage medium | |
| CN111630570A (en) | Image processing method, apparatus and computer-readable storage medium | |
| CN108028945A (en) | The apparatus and method of conversion are performed by using singleton coefficient update | |
| EP1629436B1 (en) | Overcomplete basis transform-based motion residual frame coding method and apparatus for video compression | |
| Amjoun et al. | Efficient compression of 3d dynamic mesh sequences | |
| Zhong et al. | Channel-level variable quantization network for deep image compression | |
| CN104023216A (en) | Face video compression method | |
| CN116939226A (en) | Low-code-rate image compression-oriented generated residual error repairing method and device | |
| US6594375B1 (en) | Image processing apparatus, image processing method, and storage medium | |
| Sasazaki et al. | Vector quantization of images with variable block size | |
| CN115243043A (en) | Depth image coding and decoding method and system with controllable entropy decoding complexity | |
| Ebrahimi-Moghadam et al. | Matching pursuit-based region-of-interest image coding | |
| CN115299057A (en) | Point cloud encoding method, point cloud decoding method, encoder, decoder, and storage medium | |
| CN116095321A (en) | Significant area image coding and decoding method, system, equipment and storage medium | |
| CN115442617A (en) | Video processing method and device based on video coding | |
| CN115086673A (en) | Image coding and decoding method and device for multi-code-rate and rapid entropy model calculation | |
| Kirmemis et al. | Shrinkage as activation for learned image compression | |
| Winger | Linearly constrained generalized Lloyd algorithm for reduced codebook vector quantization | |
| CN116601944A (en) | Point cloud encoding and decoding method, encoder, decoder and computer storage medium | |
| CN117750021B (en) | Video compression method, device, computer equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |