A low-complexity step-by-step refinement source coding method

Abstract

The present invention provides a low-complexity, step-by-step refinement source coding method, comprising: sparsely encoding an original source sequence using a first power allocation coefficient and a first dictionary matrix to obtain a first-layer coding result, wherein the first dictionary matrix is ​​designed based on an independent and identically distributed Gaussian codebook; decoding the first-layer coding result according to the first dictionary matrix to obtain a first source sequence; sparsely encoding the original source sequence using a second power allocation coefficient and a second dictionary matrix to obtain a second-layer coding result, wherein the second dictionary matrix is ​​designed based on an independent and identically distributed Gaussian codebook, and the distribution of the elements of the second dictionary matrix is ​​determined by the first source sequence; and decoding the second-layer coding result according to the second dictionary matrix to obtain a second source sequence. This method achieves performance similar to theoretical optimal coding with low computational and time overhead. While maintaining good performance, it adjusts the bit rate of the two-layer coding according to actual compression quality and transmission requirements, thereby providing high flexibility.

Description

Technical Field

[0001] The present invention belongs to the technical field of information source coding, and in particular relates to a low-complexity step-by-step refinement information source coding method. Background Art

[0002] Progressive refinement technology has a wide range of applications in cascaded communication systems, image and video compression, and clinical imaging diagnosis. This technology can adjust the transmission rate based on actual compression quality requirements, improving resource utilization while maintaining good performance.

[0003] Currently, stepwise refinement technology uses a two-layer encoder and decoder. Both encoders directly read and compress the source sequence, while the two decoders aim to recover the source sequence with different distortion requirements and different ways of reading the compressed information. Specifically, the first-layer encoding and decoding process involves encoder f1 encoding the source sequence S into compressed information β1, and decoder φ1 recovering the source sequence S within the distortion requirement D1 based on β1. The second-layer encoding and decoding process involves encoder f2 encoding the source sequence S into compressed information β2, and decoder φ2 using both β1 and β2 to recover the source sequence S within the more refined distortion requirement D2 (D2 < D1). The theoretically optimal design of stepwise refinement technology uses continuous minimum Euclidean distance coding. However, encoding requires sequentially traversing and calculating all possible results, resulting in a time complexity that increases exponentially with the code length, making it difficult to implement in practical applications. Summary of the Invention

[0004] In view of this, the object of the present invention is to provide a low-complexity progressive refinement source coding method to meet the demand of reducing coding complexity while achieving performance close to theoretical optimal coding.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] According to a first aspect, the present invention provides a low-complexity step-by-step refinement source coding method, comprising: obtaining an original source sequence; sparsely encoding the original source sequence using a first power allocation coefficient and a first dictionary matrix to obtain a first-layer coding result, wherein the first dictionary matrix is ​​designed based on an independent and identically distributed Gaussian codebook, and the first power allocation coefficient is limited by the power of the first-layer codebook in the step-by-step refinement coding; decoding the first-layer coding result according to the first dictionary matrix to obtain a first source sequence; sparsely encoding the original source sequence using a second power allocation coefficient and a second dictionary matrix to obtain a second-layer coding result, wherein the second dictionary matrix is ​​designed based on an independent and identically distributed Gaussian codebook, and the distribution of elements of the second dictionary matrix is ​​determined by the first source sequence, and the second power allocation coefficient is limited by the power of the second-layer codebook in the step-by-step refinement coding; decoding the second-layer coding result according to the second dictionary matrix to obtain a second source sequence.

[0007] Optionally, a low-complexity stepwise refinement source coding method, wherein the distribution of elements of the second dictionary matrix is ​​determined by the first source sequence, includes: the elements of the first dictionary matrix obey iidN(0,1); the elements of the second dictionary matrix obey iidN(μ,1), S' is the first source sequence, c 2l is the power allocation coefficient of the lth region of the second dictionary matrix.

[0008] Optionally, the first dictionary matrix is ​​a matrix with n rows and M1×L1 columns, where n is the code length. R1 is the bit rate of the first coding layer, b is the codebook parameter; the second dictionary matrix is ​​a matrix with n rows and M2×L2 columns, where R2 is the bit rate of the second coding layer,

[0009] Optionally, the original source sequence is sparsely encoded using a first power allocation coefficient and a first dictionary matrix to obtain a first-layer encoding result, including: dividing the first dictionary matrix into L1 regions, constructing an optimal solution problem for minimizing the reconstruction error based on the first power allocation coefficient and the first dictionary matrix, traversing the L1 regions, solving the optimal solution problem for minimizing the reconstruction error, and obtaining the first-layer encoding result; and / or the original source sequence is sparsely encoded using a second power allocation coefficient and a second dictionary matrix to obtain a second-layer encoding result, including: dividing the second dictionary matrix into L2 regions, constructing an optimal solution problem for minimizing the reconstruction error based on the second power allocation coefficient and the second dictionary matrix, traversing the L2 regions, solving the optimal solution problem for minimizing the reconstruction error, and obtaining the second-layer encoding result.

[0010] Optionally, constructing an optimal solution problem for minimizing the reconstruction error according to the first power allocation coefficient and the first dictionary matrix, traversing L1 regions, solving the optimal solution problem for minimizing the reconstruction error, and obtaining a first-layer coding result, including:

[0011]

[0012] Among them, 1≤i≤L1, X0=S, S represents the original source sequence, m i Indicates the position of the i-th non-zero element in the first layer encoding result, Represents the mth of the first dictionary matrix i Column, A j represents the j-th column of the first dictionary matrix, c1i is the power allocation coefficient of the ith region of the first dictionary matrix, satisfying σ 2-D1 represents the first-layer codebook power, R1 represents the code rate of the first coding layer, D1 is the distortion requirement of the first coding layer, L1 represents the number of regions in the first dictionary matrix, σ 2 Indicates the source power.

[0013] Optionally, the second dictionary matrix is ​​divided into L2 regions, an optimal solution problem for minimizing the reconstruction error is constructed according to the second power allocation coefficient and the second dictionary matrix, the L2 regions are traversed, and the optimal solution problem for minimizing the reconstruction error is solved to obtain a second-layer coding result, including:

[0014]

[0015] Among them, 1≤i≤L2, X0=S, S represents the original source sequence, n i Indicates the position of the i-th non-zero element in the second-layer encoding result, Represents the nth i Column, A j represents the j-th column of the second dictionary matrix, c 2i is the power allocation coefficient of the i-th region of the second dictionary matrix, R2 represents the bit rate of the second coding layer, D1 is the distortion requirement of the first coding layer, and L2 represents the number of regions in the second dictionary matrix.

[0016] Optionally, the size of the first dictionary matrix is ​​determined by the code rate of the first coding layer, and the size of the second dictionary matrix is ​​determined by the code rate of the second coding layer. Before using the first power allocation coefficient and the first dictionary matrix to sparsely encode the original source sequence, it includes: obtaining the first coding layer distortion requirement and the second coding layer distortion requirement; determining the code rate of the first coding layer according to the first coding layer distortion requirement and the source power, and determining the code rate of the second coding layer according to the first coding layer distortion requirement and the second coding layer distortion requirement.

[0017] Optionally, the code rate of the first coding layer is:

[0018]

[0019] Where R1 represents the bit rate of the first coding layer, σ 2 represents the source power, λ represents the code rate allocation coefficient, and D1 represents the distortion requirement of the first coding layer;

[0020]

[0021] Here, R2 represents the bit rate of the second coding layer, and D2 represents the distortion requirement of the first coding layer.

[0022] According to the second aspect, an embodiment of the present invention provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and runnable on the processor, wherein the processor executes the steps of the low-complexity stepwise refinement source coding method described in the first aspect or any embodiment of the first aspect.

[0023] According to a third aspect, an embodiment of the present invention provides a computer storage medium having computer instructions stored thereon, which, when executed by a processor, implement the steps of the low-complexity progressive refinement source coding method described in the first aspect or any embodiment of the first aspect.

[0024] An embodiment of the present invention provides a low-complexity progressive refinement source coding method, in which the first dictionary matrix and the second dictionary matrix are both designed based on independent and identically distributed Gaussian codebooks, which provides a set of over-complete basis vectors for the signal, so that the original signal can be sparsely represented on these bases, and the coefficient vectors in the coding result in the progressive refinement source coding are sparsely represented, so that the method inherits the advantages of sparse regression codes, can achieve performance close to theoretical optimal coding with low computational resource overhead and time resource overhead, and can also adjust the bit rate of the two-layer coding according to actual compression quality and transmission requirements while maintaining good performance, with high flexibility and more practical.

[0025] Other advantages, objectives and features of the present invention will be described in the following description and will be apparent to those skilled in the art to some extent, or those skilled in the art can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to make the purpose, technical solutions and beneficial effects of the present invention more clear, the present invention provides the following drawings for illustration:

[0027] Figure 1 This is a specific example flow chart of a low-complexity step-by-step refinement source coding method in the present invention;

[0028] Figure 2 This is a specific example flow chart of the present invention for obtaining a first-layer coding result by sparsely coding an original source sequence using a first power allocation coefficient and a first dictionary matrix;

[0029] Figure 3 Schematic diagram of pseudo code of the algorithm corresponding to a low-complexity step-by-step refinement source coding method of the present invention;

[0030] Figure 4 In the present invention, when the signal source power σ 2=1, when the code rate allocation coefficient λ = 1, the performance comparison diagram of the two schemes of step-by-step refinement coding proposed by this method and existing point-to-point coding under different code element transmission rates R;

[0031] Figure 5 In the present invention, when the signal source power σ 2 =1, performance comparison chart of the progressive refinement coding proposed by this method under different bit rate allocation coefficients λ;

[0032] Figure 6 In the present invention, when the signal source power σ 2 =1, the code rate allocation coefficient λ = 1, the comparison diagram of the coding performance of the progressive refinement coding proposed by this method for sources with different distributions;

[0033] Figure 7 This is a principle block diagram of a specific example of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION

[0034] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0035] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "installed," "connected," and "connected" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; internal connections between two components; wireless connections or wired connections. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0036] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0037] The embodiment of the present invention provides a low-complexity step-by-step refinement source coding method, such as Figure 1 Shown, including:

[0038] S101, obtaining the original source sequence;

[0039] S102, sparsely encoding the original source sequence using a first power allocation coefficient and a first dictionary matrix to obtain a first-layer encoding result, wherein the first dictionary matrix is ​​designed based on an independent and identically distributed Gaussian codebook, and the first power allocation coefficient is limited by the first-layer codebook power in the stepwise refinement coding;

[0040] S103, decoding the first-layer encoding result according to the first dictionary matrix to obtain a first source sequence;

[0041] S104, sparsely encoding the original source sequence using a second power allocation coefficient and a second dictionary matrix to obtain a second-layer encoding result, wherein the second dictionary matrix is ​​designed based on an independent and identically distributed Gaussian codebook, and the distribution of elements in the second dictionary matrix is ​​determined by the first source sequence, and the second power allocation coefficient is limited by the second-layer codebook power in the stepwise refinement coding;

[0042] S105 , decoding the second-layer encoding result according to the second dictionary matrix to obtain a second information source sequence.

[0043] Exemplarily, before sparsely encoding the original source sequence using the first power allocation coefficient and the first dictionary matrix, the method includes: obtaining a first coding layer distortion requirement and a second coding layer distortion requirement; determining the bit rate of the first coding layer based on the first coding layer distortion requirement and the source power, and determining the bit rate of the second coding layer based on the first coding layer distortion requirement and the second coding layer distortion requirement.

[0044] Specifically, for a given two-layer coding distortion requirement D1 and D2 (D1 < D2), the rate-distortion region of the coding is gradually refined according to the Gaussian source. Combined with the actual coding requirements, the code rate of the two layers of coding is determined accordingly and Where λ is the code rate allocation coefficient, which can be adjusted according to actual needs. The value must meet σ 2 Indicates the source power.

[0045] The first dictionary matrix A1 in the sparse recursive code is used to represent a set of basis vectors of the signal, whose elements obey iidN(0,1), and the matrix size is n×M1L1, where n is the code length. That is, the total amount of codebook is R1 is the bit rate of the first coding layer, b is a codebook parameter. In practical applications, n and b must be determined first, and b usually satisfies 1≤b≤3. Then, M1 and L1 are calculated based on R1 and the above relationship to generate the first dictionary matrix A1.

[0046]

[0047] In sparse recursive coding, the coefficient vector is a set of weights used to represent the original signal. These weights are combined with the basis vectors in the first dictionary matrix to reconstruct the original signal. That is, the goal of coding is to determine a vector β1 with M1L1 rows and 1 columns, satisfying A1β1=S, where S is the original source sequence. That is, the original source sequence is obtained by selecting several columns from the dictionary matrix A1 and adding them together. The weight of each column is determined by the first power allocation coefficient The first power allocation coefficient is determined by the power limit of the first layer codebook in the step-by-step refinement coding. Specifically, it needs to meet That is, the first layer codebook power in the step-by-step refinement coding is σ 2 -D1. Compress the original source sequence S into a coding result β1 according to the first power allocation coefficient c1 and the dictionary matrix A1.

[0048] The specific method of sparsely encoding the original source sequence using the first power allocation coefficient and the first dictionary matrix to obtain the first-layer encoding result may include: dividing the first dictionary matrix into L1 regions, constructing an optimal solution problem for minimizing the reconstruction error according to the first power allocation coefficient and the first dictionary matrix, traversing the L1 regions, solving the optimal solution problem for minimizing the reconstruction error, and obtaining the first-layer encoding result. The specific steps are as follows: Figure 2 Shown, including:

[0049] (1) Input S, c1, and A1;

[0050] (2) Initialize X0 = S, i = 1;

[0051] (3) Construct the optimal solution problem for minimizing the reconstruction error. Specifically, calculate

[0052] Among them, 1≤i≤L1, X0=S, S represents the original source sequence, m i Indicates the position of the i-th non-zero element in the first layer encoding result, Represents the mth of the first dictionary matrix i Column, A j represents the j-th column of the first dictionary matrix, c 1i is the power allocation coefficient of the ith region of the first dictionary matrix, satisfying σ 2 -D1 represents the first-layer codebook power, R1 represents the code rate of the first coding layer, D1 is the distortion requirement of the first coding layer, L1 represents the number of regions in the first dictionary matrix, σ 2 Indicates the source power.

[0053] (4) Order i=i+1;

[0054] (5) Repeat steps (3) and (4) until i = L1;

[0055] (6) For i=1,...,L1, the mth encoding result of the first layer β1 i The element is c 1i , and the rest of the elements are 0.

[0056] The first layer coding result is decoded according to the first dictionary matrix to obtain the first source sequence. Specifically, the original source sequence can be restored by multiplying the first dictionary matrix A1 and the coding result β1, S'=A1β1.

[0057] The second dictionary matrix is ​​a matrix with n rows and M2×L2 columns, where M2=L2 b and satisfy That is, the code length is the same as the first layer, and M2 and L2 are adjusted accordingly according to R2 while keeping the relationship unchanged. The elements of the second dictionary matrix obey iidN(μ,1), S' is the first source sequence, c 2l is the power allocation coefficient of the lth region of the second dictionary matrix.

[0058] The second power allocation coefficient and the second dictionary matrix are used to sparsely encode the original source sequence to obtain a second-layer encoding result, including: dividing the second dictionary matrix into L2 regions, constructing an optimal solution problem for minimizing the reconstruction error according to the second power allocation coefficient and the second dictionary matrix, traversing the L2 regions, solving the optimal solution problem for minimizing the reconstruction error, and obtaining the second-layer encoding result.

[0059] Specifically, the power of the second-layer codebook in the step-by-step refinement coding is D1-D2. Therefore, the second power allocation coefficient of the i-th column element of the second dictionary matrix is: The original source sequence S is compressed into the coding result β2 according to the second power allocation coefficient c2 and the second dictionary matrix A2. The specific coding steps are the same as steps (1) to (6) in the first layer coding, specifically:

[0060] (1) Input S, c2, and A2;

[0061] (2) Initialize X0 = S, i = 1;

[0062] (3) Construct the optimal solution problem for minimizing the reconstruction error. Specifically, calculate

[0063] Among them, 1≤i≤L2, X0=S, S represents the original source sequence, n i Indicates the position of the i-th non-zero element in the second-layer encoding result, Represents the nthi Column, A j represents the j-th column of the second dictionary matrix, c 2i is the power allocation coefficient of the i-th region of the second dictionary matrix, R2 represents the bit rate of the second coding layer, D1 is the distortion requirement of the first coding layer, and L2 represents the number of regions in the second dictionary matrix.

[0064] (4) Order i=i+1;

[0065] (5) Repeat steps (3) and (4) until i = L2;

[0066] (6) For i=1,...,L2, the nth i The element is c 2i , and the rest of the elements are 0.

[0067] The second layer coding result is decoded according to the second dictionary matrix to obtain the second source sequence. Specifically, the second dictionary matrix A2 and the second layer coding result β2 are multiplied to more accurately restore the original source sequence, S″=A2β2.

[0068] For ease of understanding, Figure 3 As shown, the pseudo code of the algorithm corresponding to the above method is given.

[0069] An embodiment of the present invention provides a low-complexity progressive refinement source coding method, in which the first dictionary matrix and the second dictionary matrix are both designed based on independent and identically distributed Gaussian codebooks, which provides a set of over-complete basis vectors for the signal, so that the original signal can be sparsely represented on these bases, and the coefficient vectors in the coding result in the progressive refinement source coding are sparsely represented, so that the method inherits the advantages of sparse regression codes, can achieve performance close to theoretical optimal coding with low computational resource overhead and time resource overhead, and can also adjust the bit rate of the two-layer coding according to actual compression quality and transmission requirements while maintaining good performance, with high flexibility.

[0070] Furthermore, in order to verify the beneficial effects of this method, the following three sets of simulation results are given for comparison:

[0071] The first set of simulation results comparison: Figure 4 As shown, it describes the situation when the source power σ 2=1, when the code rate allocation coefficient λ=1, the performance comparison of the two schemes of step-by-step refinement coding proposed by this method and existing point-to-point coding under different code element transmission rates R. The horizontal axis represents the code rate, which represents the total code rate designed according to different D2 in step-by-step refinement coding; the vertical axis represents the distortion rate, which measures the coding performance. In step-by-step refinement coding, the first-layer distortion requirement is D1=0.6, the first-layer code rate is fixed at R1=0.368 bits / transmission, and the second-layer code rate is R2=R-R1. Point-to-point source coding only performs one layer of coding, and the specific process is the same as step-by-step refinement coding. The coding code length n=605, and the dictionary matrix parameters L and M satisfy M=L 1.8 The specific value is determined by M L =2 nR The elements in the original source sequence S have a mean of 0 and a variance of σ. 2 The Gaussian distribution of . The number of repetitions per point in the simulation is 70, and other factors such as the original source sequence S of the two schemes are the same. The results show that as the bit rate increases, the distortion rate gradually decreases, and it has better performance when the bit rate is high. The progressive refinement source coding method proposed in this method is not much different from the rate-distortion function of the theoretical optimal case, but the complexity is significantly reduced. Specifically, the theoretical optimal solution needs to traverse all possible codebooks during encoding, that is, to perform M L The proposed method only requires L×M comparisons. When L and M are large, the computational complexity is far less than the theoretical optimal solution, making it highly feasible and easy to apply in practical systems. Although the performance of progressive refinement coding is slightly inferior to that of point-to-point coding, it has high application value. For example, during medical imaging examinations, doctors usually first browse the image as a whole. At this time, the image accuracy requirements are not high. This corresponds to the first layer of encoding in progressive refinement coding, which encodes at a lower bit rate at a relatively low compression quality, achieving the goal with minimal resource consumption. When a suspected lesion is found in a certain area, a more accurate image of this area is required. At this time, a lower distortion rate is required. This corresponds to the second layer of encoding in progressive refinement coding, which meets higher compression quality. Therefore, progressive refinement coding can allocate resources according to actual needs and improve resource utilization.

[0072] The second set of simulation results comparison: Figure 5 As shown, it describes the situation when the source power σ 2 =1, the performance comparison of the progressive refinement coding proposed by this method under different code rate allocation coefficients λ. The horizontal axis represents the total code rate R, and the vertical axis represents the distortion rate. The first layer distortion requirement is D1 = 0.7, the first layer code rate The second layer code rate R2 = R-R1. The encoding code length n = 605, the dictionary matrix parameters L and M satisfy M = L 1.8 The specific value is determined by M L =2 nRCalculated. The elements in the source sequence S have a mean of 0 and a variance of σ 2 The Gaussian distribution of the two layers is used. Each point is repeated 70 times during simulation, while other factors, such as the original source sequence S, remain constant. Results show that the performance of the progressive refinement coding approach is similar when the rate allocation coefficient λ is varied, with a larger λ resulting in better performance. Therefore, in general, setting λ = 1 offers superior performance. Furthermore, adjusting the rate allocation between the two layers of coding using λ can be used to adjust the code rate based on actual needs. This results in minimal performance change, but can better meet practical requirements and has significant practical implications.

[0073] The third group of simulation results comparison: Figure 6 As shown, it describes the situation when the source power σ 2 =1, and the code rate allocation coefficient λ =1, the coding performance comparison of the progressive refinement coding proposed in this method for sources with different distributions. The horizontal axis represents the total code rate R, and the vertical axis represents the distortion rate. The first-layer distortion requirement is D1 = 0.6, the first-layer code rate is fixed R1 = 0.368 bits / transmission, and the second-layer code rate R2 = R-R1. The encoding code length n = 605, and the dictionary matrix parameters L and M satisfy M = L 1.8 The specific value is calculated based on M L =2 nR The simulation was repeated 70 times per point. The results show that the performance of the progressive refinement coding approach is similar for sources with different distributions. This method demonstrates robust coding and decoding, and the coding dictionary matrix design is independent of the source distribution, resulting in equally good coding performance for sources with different distributions.

[0074] The present application also provides an electronic device, such as Figure 7 As shown, a processor 501 and a memory 502 , wherein the processor 501 and the memory 502 may be connected via a bus or other means.

[0075] The processor 501 may be a central processing unit (CPU). The processor 501 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or a combination of the above chips.

[0076] Memory 502, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the low-complexity stepwise refinement source coding method in the embodiments of the present invention. The processor executes the non-transitory software programs, instructions, and modules stored in the memory to perform various processor functions and data processing.

[0077] The memory 502 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created by the processor, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-transitory memory, such as at least one disk storage device, a flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory 502 may optionally include a memory remotely located relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0078] The one or more modules are stored in the memory 502 and when executed by the processor 501, perform the following steps: Figure 1 The low-complexity step-by-step refinement source coding method in the illustrated embodiment.

[0079] For details of the above electronic equipment, please refer to Figure 1 The corresponding descriptions and effects in the embodiments shown can be understood and will not be repeated here.

[0080] This embodiment further provides a computer storage medium storing computer-executable instructions capable of executing the low-complexity step-by-step refinement source coding method of any of the above-described method embodiments. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD), or a solid-state drive (SSD); the storage medium may also include a combination of the above-described types of memory.

[0081] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made in form and details without departing from the scope defined by the claims of the present invention.

Claims

1. A low-complexity step-by-step refinement source coding method, characterized in that: include: Obtain the original source sequence; The original source sequence is sparsely coded using a first power allocation coefficient and a first dictionary matrix to obtain a first-layer coding result. The first dictionary matrix is ​​designed based on an independent and identically distributed Gaussian codebook, and the first power allocation coefficient is limited by the power of the first-layer codebook in the stepwise refinement coding. Decoding the first layer encoding result according to the first dictionary matrix to obtain a first source sequence; The original source sequence is sparsely coded using a second power allocation coefficient and a second dictionary matrix to obtain a second-layer coding result. The second dictionary matrix is ​​designed based on an independent and identically distributed Gaussian codebook, and the distribution of the elements of the second dictionary matrix is ​​determined by the first source sequence. The second power allocation coefficient is limited by the second-layer codebook power in the stepwise refinement coding. Decoding the second layer encoding result according to the second dictionary matrix to obtain a second source sequence; The distribution of elements of the second dictionary matrix is ​​determined by the first source sequence, including: The elements of the first dictionary matrix obey iidN(0,1); The elements of the second dictionary matrix obey iidN( ,1), , is the first source sequence, is the first Power allocation coefficient for each region; The first dictionary matrix has n rows and n columns. , where n is the code length, , is the bit rate of the first coding layer, , b is the codebook parameter; The second dictionary matrix has n rows and n columns. The matrix, where , is the bit rate of the second coding layer, ; The first power allocation coefficient and the first dictionary matrix are used to sparsely encode the original source sequence to obtain the first layer encoding result, including: dividing the first dictionary matrix into regions, construct the optimal solution for minimizing the reconstruction error based on the first power allocation coefficient and the first dictionary matrix, and traverse Regions, solve the problem of minimizing the optimal solution of reconstruction error, and obtain the first layer coding result; The second power allocation coefficient and the second dictionary matrix are used to sparsely encode the original source sequence to obtain the second layer encoding result, including: dividing the second dictionary matrix into regions, construct the optimal solution for minimizing the reconstruction error based on the second power allocation coefficient and the second dictionary matrix, and traverse Regions, solve the problem of minimizing the reconstruction error to obtain the second-layer coding result; According to the first power allocation coefficient and the first dictionary matrix, the optimal solution for minimizing the reconstruction error is constructed. Regions, solve the problem of minimizing the reconstruction error to obtain the optimal solution, and obtain the first-layer coding results, including: ； in, , , , represents the original source sequence, Indicates the position of the i-th non-zero element in the first layer encoding result, The first dictionary matrix List, represents the j-th column of the first dictionary matrix, , is the power allocation coefficient of the ith region of the first dictionary matrix, satisfying , represents the first layer codebook power, represents the bit rate of the first coding layer, is the first coding layer distortion requirement, Indicates the number of regions in the first dictionary matrix, Indicates the source power; Divide the second dictionary matrix into regions, construct the optimal solution for minimizing the reconstruction error based on the second power allocation coefficient and the second dictionary matrix, and traverse Regions, solve the problem of minimizing the reconstruction error to obtain the optimal solution, and obtain the second-layer coding results, including: ； in, , , , represents the original source sequence, Indicates the position of the i-th non-zero element in the second-layer encoding result, The first List, represents the j-th column of the second dictionary matrix, , is the power allocation coefficient of the ith region of the second dictionary matrix, represents the bit rate of the second coding layer, is the first coding layer distortion requirement, Indicates the number of regions in the second dictionary matrix; The size of the first dictionary matrix is ​​determined by the code rate of the first coding layer, and the size of the second dictionary matrix is ​​determined by the code rate of the second coding layer. Before sparse coding the original source sequence using the first power allocation coefficient and the first dictionary matrix, the method includes: Obtaining a first coding layer distortion requirement and a second coding layer distortion requirement; Determine a bit rate of the first coding layer according to the distortion requirement of the first coding layer and the signal source power, and determine a bit rate of the second coding layer according to the distortion requirement of the first coding layer and the distortion requirement of the second coding layer; The code rate of the first coding layer is: ； in, represents the bit rate of the first coding layer, represents the source power, represents the code rate allocation coefficient, Indicates the distortion requirement of the first coding layer; ； in, represents the bit rate of the second coding layer, Indicates the distortion requirement of the first coding layer.

2. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the steps of the low-complexity progressive refinement source coding method according to claim 1.

3. A computer storage medium having computer instructions stored thereon, characterized in that: When the instruction is executed by a processor, the steps of the low-complexity stepwise refinement source coding method according to claim 1 are implemented.

Images