A method for optimizing design of a compressed imaging block circulant structured measurement matrix

Abstract

The application discloses a kind of compressed imaging block circulation structure measurement matrix optimization design method, belong to image technical field.The method includes: based on the total pixel number of image to be reconstructed constructs particle swarm, based on the current free element vector of each particle determines the current block circulation measurement matrix, calculates the Gram matrix corresponding to the current block circulation measurement matrix;Using the objective function based on Gram matrix constructs to calculate the fitness of each particle;Based on the current free element vector and fitness of each particle determines and updates the local optimal free element vector corresponding to each particle, determines the inertia weight of each particle;From all local optimal free element vectors, determine the global optimal free element vector;Based on global optimal free element vector determines the block circulation measurement matrix corresponding to the compressed sampling, reconstructs image based on block circulation measurement matrix.The application can improve subsequent image reconstruction effect.

Description

Technical Field

[0001] This invention belongs to the field of image technology, and in particular relates to an optimization design method for a block loop structure measurement matrix in compressed imaging. Background Technology

[0002] Compressed imaging is a revolutionary imaging technique that breaks through the traditional Nyquist sampling theorem. Its core idea is to utilize the sparsity of the signal to reconstruct the original signal from a much smaller number of random linear measurements through nonlinear projection. This technique greatly alleviates the pressure on imaging systems in terms of sampling, transmission, and storage, and therefore has wide application value in high-resolution remote sensing imaging, multispectral optical imaging, high-resolution infrared imaging, and high-speed photography. Currently, the key challenges of compressed imaging technology lie in dedicated physical hardware implementation and efficient image reconstruction algorithms.

[0003] The design of the measurement matrix is ​​crucial for the physical realization of compressed imaging. The core challenge in measurement matrix design lies in the trade-off between theoretical optimality and hardware feasibility. For example, a Gaussian random matrix possesses theoretical optimality because it satisfies the finite isometry property with a relatively high probability. However, its large number of free elements hinders hardware implementation. While structured measurement matrices (such as block cyclic matrices) are easy to implement in hardware, their finite isometry property, satisfied with only a certain probability, leads to poor data reconstruction results. Therefore, optimizing the design of structured measurement matrices is an effective way to alleviate this contradiction. Most existing methods are based on deterministic iterative optimization of Gram matrices. These methods are computationally expensive (especially for large-scale matrices) and easily disrupt the matrix structure when used for block cyclic measurement matrices. Therefore, optimizing the design of structured measurement matrices such as block cyclic measurement matrices is an effective approach for the physical realization of compressed imaging systems. Summary of the Invention

[0004] This invention proposes an optimized design method for the block-loop structure measurement matrix of compressed imaging, which solves the technical problems of complex hardware implementation, large-scale storage and low robustness in compressed imaging engineering.

[0005] The first aspect of this invention proposes a method for optimizing the design of a block loop structure measurement matrix for compressed imaging, the method comprising:

[0006] Step S1: Denote the total number of pixels in the image to be reconstructed as n;

[0007] Step S2: Construct a particle swarm containing N particles, each particle being a randomly generated 1×n free element vector and a 1×n velocity vector, wherein the free element vector is used to construct the block cyclic measurement matrix; initialize the current iteration number d to 1;

[0008] Step S3: Determine the current block cycle measurement matrix based on the current free element vectors of each particle, and calculate the Gram matrix corresponding to the current block cycle measurement matrix; use the objective function constructed based on the Gram matrix to calculate the fitness of each particle;

[0009] Based on the current free element vectors and fitness of each particle, determine and update the local optimal free element vectors corresponding to each particle, and determine the inertia weights of each particle; determine the global optimal free element vector from all local optimal free element vectors.

[0010] Based on the current free element vector, current velocity vector, and inertia weight of each particle, determine the current free element vector and current velocity vector of each particle when the current iteration number is d+1;

[0011] Step S4: If the termination condition is not met, assign d to d+1 and proceed to step S3; otherwise, proceed to step S5.

[0012] Step S5: Determine the block cyclic measurement matrix corresponding to the sample to be compressed based on the globally optimal free element vector, and reconstruct the image based on the block cyclic measurement matrix.

[0013] Preferably, in step S2, a particle swarm containing N particles is constructed, where each particle is a randomly generated free element vector of dimension 1×n and a velocity vector of dimension 1×n, including:

[0014] The total number of pixels in the image to be reconstructed is n = r × c, where r and c are the number of rows and columns of the image to be reconstructed, respectively; then the compression sampling ratio CR = m / n, where m is the number of samples; the size of the block cyclic measurement matrix corresponding to the compression sampling is determined to be m × n;

[0015] Initialize based on the size of the block loop measurement matrix. N The particle swarm consists of free element vectors with dimensions of 1×n. Each particle is initialized with a 1×n vector as its velocity vector, and all velocity vectors form a set. Where N is the number of particles in the particle swarm. The first in the particle swarm i One particle, It is the first i The velocity vector of each particle;

[0016] The size of the sub-matrix block of the block cyclic measurement matrix is ​​determined to be q×q, where q is a common divisor of m and n. The number of sub-matrix blocks is determined to be k = n / q.

[0017] Preferably, in step S3, determining the current block cycle measurement matrix based on the current free element vectors of each particle, and calculating the Gram matrix corresponding to the current block cycle measurement matrix, includes:

[0018] Calculate the product of the current block loop measurement matrix and the sparse transformation matrix required for image reconstruction to obtain the matrix. , For block cyclic measurement matrix, It is a sparse transformation matrix. It is the set of real numbers;

[0019] Calculate matrix Column normalized matrix The Gram matrix corresponding to the current block cyclic measurement matrix is: ,and .

[0020] Preferably, in step S3, the objective function J is:

[0021]

[0022]

[0023]

[0024] Where H is the target matrix, For the Welch community, s , t These are the row and column values ​​of the elements of the target matrix H, respectively. It is the Frobenius norm. For the target matrix H, the first s Okay, number t Column elements, For Gram matrix G, the first... s Okay, number t The number of columns in the block loop measurement matrix is ​​n, the number of rows in the block loop measurement matrix is ​​m, and the sign function is the sign function.

[0025] Preferably, in step S3, the formula for calculating the inertia weight is:

[0026]

[0027]

[0028] in, Let be the inertia weight of the i-th particle when the current iteration number is d. A constant between (0,1] The coefficient representing an individual's ability to find the best. This refers to the i-th particle when the current iteration number is d. Let be the local optimal free element vector of the i-th particle when the current iteration number is d. This represents the globally optimal free element vector when the current iteration number is d. It is a positive number, used to prevent the denominator from being 0.

[0029] Preferably, in step S3, based on the current free element vector, current velocity vector, and inertia weight of each particle, the current free element vector and current velocity vector of each particle when the current iteration number is d+1 are determined, wherein:

[0030]

[0031]

[0032] in, , These are the first and second random numbers, respectively, which are between [0,1]. , These are the first learning factor and the second learning factor, respectively. and Let be the particle and its current velocity vector at the current iteration number d+1. and This represents the particle and its current velocity vector at the current iteration number d. Let be the inertial weight of the i-th particle when the current iteration number is d.

[0033] Preferably, the termination condition in step S4 is:

[0034] in, Let be the globally optimal free element vector when the current iteration number is d, and let be the globally optimal free element vector when the current iteration number is d-1; when d-1 equals 0... The initialized global optimal free element vector. For the set threshold, Norm operations.

[0035] Preferably, step S5, determining the block cyclic measurement matrix corresponding to the sample to be compressed based on the globally optimal free element vector, includes:

[0036] Divide all elements of the global optimal free element vector into k parts in order, each part containing q elements; use the q elements in each part to generate a sub-circular matrix block, and then combine each sub-circular matrix block into the block circular measurement matrix corresponding to the sample to be compressed.

[0037] A second aspect of the present invention provides a device for optimizing the design of a block loop structure measurement matrix for compressed imaging, the device comprising:

[0038] Information acquisition module: configured to record the total number of pixels in the image to be reconstructed as n;

[0039] Initialization module: Configured to build a particle swarm containing N particles, each particle being a randomly generated 1×n free element vector and a 1×n velocity vector, where the free element vector is used to construct the block loop measurement matrix; initialize the current iteration number d to 1;

[0040] Iteration module: Configured to determine the current block cycle measurement matrix based on the current free element vectors of each particle, calculate the Gram matrix corresponding to the current block cycle measurement matrix, and calculate the fitness of each particle using the objective function constructed based on the Gram matrix;

[0041] Based on the current free element vectors and fitness of each particle, determine and update the local optimal free element vectors corresponding to each particle, and determine the inertia weights of each particle; determine the global optimal free element vector from all local optimal free element vectors.

[0042] Based on the current free element vector, current velocity vector, and inertia weight of each particle, determine the current free element vector and current velocity vector of each particle when the current iteration number is d+1;

[0043] The judgment module is configured to assign the value d+1 to d if the termination condition is not met, and trigger the iteration module; otherwise, it will trigger the reconstruction module.

[0044] Reconstruction module: Configured to determine the block cyclic measurement matrix corresponding to the sample to be compressed based on the globally optimal free element vector, and reconstruct the image based on the block cyclic measurement matrix.

[0045] A third aspect of the present invention provides an electronic device, the electronic device comprising:

[0046] At least one processor; and

[0047] A memory communicatively connected to the at least one processor; wherein,

[0048] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method as described above.

[0049] A fourth aspect of the present invention provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform the method described above.

[0050] Based on the coherence discrimination criterion, this invention constructs a target matrix by using Welch bounds as threshold constraints on the off-diagonal elements of the Gram matrix, and establishes an objective function by making the Gram matrix approximate the target matrix. Then, using the free element vectors of the constructed block cyclic measurement matrix as the optimization object, the objective function is minimized using a particle swarm optimization algorithm to optimize the block cyclic measurement matrix. This invention can reduce its correlation with the sparse transformation matrix while maintaining the structure of the block cyclic measurement matrix, thereby improving the subsequent image reconstruction effect.

[0051] The present invention has the following technical effects:

[0052] (1) This application can overcome the problems of traditional deterministic iterative optimization methods that easily destroy the measurement matrix structure and the high computational cost brought about by large-scale matrix operations. It achieves a balance between the theoretical optimality of the measurement matrix and the hardware feasibility.

[0053] (2) This application effectively solves the optimization problem of block cyclic measurement matrix by using particle swarm optimization method based on correlation discrimination criterion. Since this technique takes the free element vector of the constructed matrix as the optimization object and uses the Welch bound of the correlation coefficient as the threshold constraint of the off-diagonal elements of the Gram matrix to construct the objective function, it achieves high-performance nonlinear projection effect of block cyclic measurement matrix.

[0054] (3) The block cyclic measurement matrix optimization design method provided in this application is also of reference value for the optimization of other structural measurement matrices, such as binary block cyclic measurement matrices and Toeplitz measurement matrices. The process only requires improvement of the initialization of the particle swarm. Therefore, the method provided in this application has broad application value for the optimization of structural measurement matrices in compressed imaging. Attached Figure Description

[0055] Figure 1 A flowchart illustrating the block loop structure measurement matrix optimization design method for compressed imaging provided by this invention. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0057] Optical compressed imaging achieves compressed sampling by adding an optical modulator to the optical system to modulate optical parameters. This optical modulator has a block loop structure. For example... Figure 1 As shown, a method for optimizing the design of a block loop structure measurement matrix for compressed imaging is described. The method includes:

[0058] Step S1: Denote the total number of pixels in the image to be reconstructed as n;

[0059] Step S2: Construct a particle swarm containing N particles, each particle being a randomly generated 1×n free element vector and a 1×n velocity vector, wherein the free element vector is used to construct the block cyclic measurement matrix; initialize the current iteration number d to 1;

[0060] Step S3: Determine the current block cycle measurement matrix based on the current free element vectors of each particle, and calculate the Gram matrix corresponding to the current block cycle measurement matrix; use the objective function constructed based on the Gram matrix to calculate the fitness of each particle;

[0061] Based on the current free element vectors and fitness of each particle, determine and update the local optimal free element vectors corresponding to each particle, and determine the inertia weights of each particle; determine the global optimal free element vector from all local optimal free element vectors.

[0062] Based on the current free element vector, current velocity vector, and inertia weight of each particle, determine the current free element vector and current velocity vector of each particle when the current iteration number is d+1;

[0063] Step S4: If the termination condition is not met, assign d to d+1 and proceed to step S3; otherwise, proceed to step S5.

[0064] Step S5: Determine the block cyclic measurement matrix corresponding to the sample to be compressed based on the globally optimal free element vector, and reconstruct the image based on the block cyclic measurement matrix.

[0065] Further, in step S2, a particle swarm containing N particles is constructed, where each particle is a randomly generated free element vector of dimension 1×n and a velocity vector of dimension 1×n, including:

[0066] The total number of pixels in the image to be reconstructed is n = r × c, where r and c are the number of rows and columns of the image to be reconstructed, respectively; then the compression sampling ratio CR = m / n, where m is the number of samples; the size of the block cyclic measurement matrix corresponding to the compression sampling is determined to be m × n;

[0067] Initialize based on the size of the block loop measurement matrix. N The particle swarm consists of free element vectors with dimensions of 1×n. Each particle is initialized with a 1×n vector as its velocity vector, and all velocity vectors form a set. Where N is the number of particles in the particle swarm. The first in the particle swarm i One particle, It is the first i The velocity vector of each particle;

[0068] The size of the sub-matrix block of the block cyclic measurement matrix is ​​determined to be q×q, where q is a common divisor of m and n. The number of sub-matrix blocks is determined to be k = n / q.

[0069] In this invention, different free element vectors can be used to construct corresponding block cyclic measurement matrices according to the mathematical expression of the block cyclic measurement matrix. For example, in a 128×256 block cyclic measurement matrix, where the submatrix size is 16×16, the number of elements in the free element vectors that construct this block cyclic measurement matrix is ​​16×(256 / 16)=256.

[0070] For example, block cyclic measurement matrix The mathematical form can be expressed as

[0071]

[0072] and,

[0073] in, For block cyclic measurement matrix, Let i be the i-th sub-block matrix in the block cyclic measurement matrix. It also has a cyclic structure. k = n / q , l = m / q , l It is an intermediate variable. Middle elements The elements are taken from the free element vector. If the optimization target is a block cyclic measurement matrix of size 128×256, where the submatrix size is 16×16, then the number of free elements constructing this block cyclic measurement matrix is ​​16×(256 / 16)=256. Therefore, Each element in the vector is a vector with dimension 256.

[0074] In step S3, the current block cycle measurement matrix is ​​determined based on the current free element vectors of each particle, and the Gram matrix corresponding to the current block cycle measurement matrix is ​​calculated, including:

[0075] Calculate the product of the current block cyclic measurement matrix and the sparse transformation matrix required for image reconstruction to obtain matrix D;

[0076] Calculate the product of the current block loop measurement matrix and the sparse transformation matrix required for image reconstruction to obtain the matrix. , For block cyclic measurement matrix, It is a sparse transformation matrix. It is the set of real numbers.

[0077] In this invention, the sparse transformation matrix is ​​the DCT transformation matrix. Based on the correlation discrimination criterion, the block cyclic measurement matrix is ​​reduced. With sparse transformation matrix The correlation between the two can also optimize their performance. The correlation between the two is equivalent to a matrix. The maximum value of the normalized cross-correlation coefficients between the columns of matrix D. If the column-normalized matrix of matrix D is represented as... Then the Gram matrix is The sparse transformation matrix here can be a sparse transformation basis or an overcomplete redundant dictionary.

[0078] Furthermore, in step S3, the objective function J is:

[0079]

[0080]

[0081]

[0082] Where H is the target matrix, For the Welch community, s , t These are the row and column values ​​of the elements of the target matrix H, respectively. It is the Frobenius norm. For the target matrix H, the first s Okay, number t Column elements, For Gram matrix G, the first... s Okay, number t The number of columns in the block loop measurement matrix is ​​n, the number of rows in the block loop measurement matrix is ​​m, and the sign function is the sign function.

[0083] In this invention, the goal of measurement matrix optimization is to minimize the maximum cross-correlation coefficient so that all projected vectors contain the same redundancy. Measurement matrix With sparse transformation matrix The cross-correlation coefficients between the measurement matrix and the sparse transformation matrix have a lower bound, known as the Welch bound. By approximating the Welch bound to the cross-correlation coefficients between the measurement matrix and the sparse transformation matrix, each measurement value can contain an equal amount of information, thus maximizing the information entropy of the measurement value. To improve the convergence speed, this invention uses the Welch bound threshold to truncate the off-diagonal elements of the Gram matrix, defining them as the target matrix H. This invention constrains the off-diagonal elements of the H matrix while setting the diagonal elements of the H matrix to 1. An optimization objective function is constructed by approximating the target matrix H to the Gram matrix. By minimizing the objective function, the optimal block cyclic measurement matrix is ​​obtained.

[0084] As the number of iterations increases, the fitness of each particle is recorded at each iteration, thereby determining the local optimal free element vector for each particle. The global optimal free element vector is determined from all local optimal free element vectors. The local optimal free element vector and the global optimal free element vector for each particle are continuously updated.

[0085] Furthermore, in step S3, the formula for calculating the inertia weight is:

[0086]

[0087]

[0088] in, Let be the inertia weight of the i-th particle when the current iteration number is d. A constant between (0,1] The coefficient representing an individual's ability to find the best. For the i-th particle when the current iteration number is d, Let be the local optimal free element vector of the i-th particle when the current iteration number is d. This represents the globally optimal free element vector when the current iteration number is d. It is a positive number.

[0089] In this invention, The value is usually 0.3. The value is a positive number close to 0 to prevent the denominator from being zero and causing computational failure. In the optimization of a block cyclic measurement matrix of size 128×256, the sparse transformation matrix is ​​a DCT matrix, the number of particles is 50, the particle element values ​​are constrained to the interval [-10, 10], and the objective function value changes with the number of iterations.

[0090] Further, in step S3, based on the current free element vector, current velocity vector, and inertia weight of each particle, the current free element vector and current velocity vector of each particle when the current iteration number is d+1 are determined, wherein:

[0091]

[0092]

[0093] in, , These are the first and second random numbers, respectively, which are between [0,1]. , These are the first learning factor and the second learning factor, respectively. and Let be the particle and its current velocity vector at the current iteration number d+1. and This represents the particle and its current velocity vector at the current iteration number d. Let be the inertial weight of the i-th particle when the current iteration number is d.

[0094] The termination condition in step S4 is:

[0095] in, Let be the globally optimal free element vector when the current iteration number is d, and let be the globally optimal free element vector when the current iteration number is d-1; when d-1 equals 0... The initialized global optimal free element vector. For the set threshold, Norm operations.

[0096] In this invention, This is a positive number close to 0, set as needed. In the first 100 iterations, the elements in the particle change drastically, corresponding to a rapid decrease in fitness value. From iterations 100 to 200, the changes in elements become relatively gradual, corresponding to a gradual convergence of fitness values. Finally, although a few elements remain in a changing state, their impact on the fitness of the global optimum is minimal. At this point, the termination condition is met, resulting in a vector containing 256 free elements. In practical applications, the selection of parameters and the termination condition can be set as needed.

[0097] Step S5, determining the block cyclic measurement matrix corresponding to the sample to be compressed based on the globally optimal free element vector, includes:

[0098] Divide all elements of the global optimal free element vector into k parts in order, each part containing q elements; use the q elements in each part to generate a sub-circular matrix block, and then combine each sub-circular matrix block into the block circular measurement matrix corresponding to the sample to be compressed.

[0099] In this invention, a corresponding block cyclic measurement matrix is ​​constructed according to the mathematical expression of the block cyclic matrix. Thus, this application takes the free element vector of the block cyclic measurement matrix as the optimization object, constructs the objective function according to the correlation discrimination criterion, and uses the particle swarm optimization algorithm for iterative calculation to complete the optimized design of the block cyclic measurement matrix.

[0100] For example, the size of the block cyclic matrix is ​​6. 9, meaning m=6, n=9. The size of the sub-block matrix in the block cyclic matrix is ​​3. 3, i.e., q=3. Therefore, each row of the block cyclic matrix has n / q=3 sub-block matrices, and each column has m / q=2 sub-block matrices. Assuming the final free element vector is [1,2,3,4,5,6,7,8,9], we first construct 3 sub-block matrices by dividing these 9 elements into three groups: 1,2,3, 4,5,6,7,8,9. We then iterate through each row by shifting one element to the right, resulting in three sub-block matrices.

[0101]

[0102] Then, a block circular matrix is ​​constructed from the three sub-blocks.

[0103]

[0104] The apparatus provided for carrying out the present invention will be described below. The specific implementation process and technical effects are as described above and will not be repeated below.

[0105] Optionally, embodiments of the present invention provide a device for optimizing the design of a block loop structure measurement matrix for compressed imaging, the device comprising:

[0106] Information acquisition module: configured to record the total number of pixels in the image to be reconstructed as n;

[0107] Initialization module: Configured to build a particle swarm containing N particles, each particle being a randomly generated 1×n free element vector and a 1×n velocity vector, where the free element vector is used to construct the block loop measurement matrix; initialize the current iteration number d to 1;

[0108] Iteration module: Configured to determine the current block cycle measurement matrix based on the current free element vectors of each particle, calculate the Gram matrix corresponding to the current block cycle measurement matrix, and calculate the fitness of each particle using the objective function constructed based on the Gram matrix;

[0109] Based on the current free element vectors and fitness of each particle, determine and update the local optimal free element vectors corresponding to each particle, and determine the inertia weights of each particle; determine the global optimal free element vector from all local optimal free element vectors.

[0110] Based on the current free element vector, current velocity vector, and inertia weight of each particle, determine the current free element vector and current velocity vector of each particle when the current iteration number is d+1;

[0111] The judgment module is configured to assign the value d+1 to d if the termination condition is not met, and trigger the iteration module; otherwise, it will trigger the reconstruction module.

[0112] Reconstruction module: Configured to determine the block cyclic measurement matrix corresponding to the sample to be compressed based on the globally optimal free element vector, and reconstruct the image based on the block cyclic measurement matrix.

[0113] The above-described device is used to execute the method provided in the foregoing embodiments, and its implementation principle and technical effect are similar, so they will not be described again here.

[0114] These modules can be one or more integrated circuits configured to implement the above methods, such as one or more Application Specific Integrated Circuits (ASICs), one or more digital signal processors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs). Alternatively, when a module is implemented using processing element scheduler code, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. Furthermore, these modules can be integrated together as a system-on-a-chip (SOC).

[0115] The modules described above can be connected or communicate with each other via wired or wireless connections. Wired connections may include metal cables, optical fibers, hybrid cables, or any combination thereof. Wireless connections may include connections via LAN, WAN, Bluetooth, ZigBee, or NFC, or any combination thereof. Two or more modules can be combined into a single module, and any module can be divided into two or more units. Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems and devices described above can be referred to the corresponding processes in the method embodiments, and will not be repeated here.

[0116] It should be noted that these modules can be one or more integrated circuits configured to implement the above methods, such as one or more Application Specific Integrated Circuits (ASICs), one or more Digital Signal Processors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs). Furthermore, when a module is implemented using processing element scheduler code, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. Additionally, these modules can be integrated together to form a System-on-a-Chip (SOC).

[0117] The electronic device includes a processor, memory, communication interface, display screen, and input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, Near Field Communication (NFC), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input device can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the device's casing, or an external keyboard, touchpad, or mouse.

[0118] The present invention also provides a program product, such as a computer-readable storage medium, including a program that, when executed by a processor, is used to perform the above-described method embodiments.

[0119] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0120] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0121] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0122] The integrated units implemented as software functional units described above can be stored in a computer-readable storage medium. These software functional units, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for optimizing the design of a block loop structure measurement matrix for compressed imaging, characterized in that, The methods include: Step S1: Denote the total number of pixels in the image to be reconstructed as n; Step S2: Construct a particle swarm containing N particles, each particle being a randomly generated 1×n free element vector and a 1×n velocity vector, wherein the free element vector is used to construct the block cyclic measurement matrix; initialize the current iteration number d to 1; Step S3: Determine the current block cycle measurement matrix based on the current free element vectors of each particle, and calculate the Gram matrix corresponding to the current block cycle measurement matrix; use the objective function constructed based on the Gram matrix to calculate the fitness of each particle; Based on the current free element vectors and fitness of each particle, determine and update the local optimal free element vectors corresponding to each particle, determine the inertia weights of each particle, and determine the global optimal free element vector from all local optimal free element vectors; Based on the current free element vector, current velocity vector, and inertia weight of each particle, determine the current free element vector and current velocity vector of each particle when the current iteration number is d+1; Step S4: If the termination condition is not met, assign d to d+1 and proceed to step S3; otherwise, proceed to step S5. Step S5: Determine the block cyclic measurement matrix corresponding to the sample to be compressed based on the globally optimal free element vector, and reconstruct the image based on the block cyclic measurement matrix.

2. The method as described in claim 1, characterized in that, In step S2, a particle swarm containing N particles is constructed. Each particle is a randomly generated free element vector of dimension 1×n and a velocity vector of dimension 1×n, including: The total number of pixels in the image to be reconstructed is n = r × c, where r and c are the number of rows and columns of the image to be reconstructed, respectively; then the compression sampling ratio CR = m / n, where m is the number of samples; the size of the block cyclic measurement matrix corresponding to the compression sampling is determined to be m × n; Initialize based on the size of the block loop measurement matrix. N The particle swarm consists of free element vectors with dimensions of 1×n. Each particle is initialized with a 1×n vector as its velocity vector, and all velocity vectors form a set. Where N is the number of particles in the particle swarm. The first in the particle swarm i One particle, It is the first i The velocity vector of each particle; The size of the sub-matrix block of the block cyclic measurement matrix is ​​determined to be q×q, where q is a common divisor of m and n. The number of sub-matrix blocks is determined to be k = n / q.

3. The method as described in claim 2, characterized in that, In step S3, the current block cycle measurement matrix is ​​determined based on the current free element vectors of each particle, and the Gram matrix corresponding to the current block cycle measurement matrix is ​​calculated, including: Calculate the product of the current block loop measurement matrix and the sparse transformation matrix required for image reconstruction to obtain the matrix. , For block cyclic measurement matrix, It is a sparse transformation matrix. It is the set of real numbers; Calculate matrix Column normalized matrix The Gram matrix corresponding to the current block cyclic measurement matrix is: ,and .

4. The method as described in claim 3, characterized in that, In step S3, the objective function J is: Where H is the target matrix, For the Welch community, s , t These are the row and column values ​​of the elements of the target matrix H, respectively. It is the Frobenius norm. For the target matrix H, the first s line, number t Column elements, For Gram matrix G, the first... s line, number t The number of columns in the block loop measurement matrix is ​​n, the number of rows in the block loop measurement matrix is ​​m, and the sign function is the sign function.

5. The method as described in claim 4, characterized in that, In step S3, the formula for calculating the inertia weight is: in, Let be the inertia weight of the i-th particle when the current iteration number is d. A constant between (0,1] The coefficient representing an individual's ability to find the best. This refers to the i-th particle when the current iteration number is d. Let be the local optimal free element vector of the i-th particle when the current iteration number is d. This is the globally optimal free element vector when the current iteration number is d; It is a positive number, used to prevent the denominator from being 0.

6. The method as described in claim 5, characterized in that, In step S3, based on the current free element vector, current velocity vector, and inertia weight of each particle, the current free element vector and current velocity vector of each particle when the current iteration number is d+1 are determined, wherein: in, , These are the first and second random numbers, respectively, which are between [0,1]. , These are the first learning factor and the second learning factor, respectively. and Let be the particle and its current velocity vector at the current iteration number d+1. and This represents the particle and its current velocity vector at the current iteration number d. Let be the inertial weight of the i-th particle when the current iteration number is d.

7. The method as described in claim 6, characterized in that, The termination condition in step S4 is: Where, when d-1 equals 0, The initialized global optimal free element vector. For the set threshold, Norm operations.

8. The method as described in claim 7, characterized in that, Step S5, determining the block cyclic measurement matrix corresponding to the sample to be compressed based on the globally optimal free element vector, includes: Divide all elements of the global optimal free element vector into k parts in order, each part containing q elements; use the q elements in each part to generate a sub-circular matrix block, and then combine each sub-circular matrix block into the block circular measurement matrix corresponding to the sample to be compressed.

Images