A cup-visar compression image inversion method and device

Abstract

The application discloses a CUP-VISAR compressed image inversion method and equipment, and the CUP-VISAR compressed image inversion method comprises the following steps: obtaining a two-dimensional fringe compressed image, and expressing the two-dimensional fringe compressed image by using a mathematical equation; inversely solving the mathematical equation to invert the two-dimensional fringe compressed image, and obtaining an original interference fringe image; and verifying the feasibility of the inversion method. The application has the effect of improving image quality.

Description

Technical Field

[0001] This invention belongs to the field of image processing, specifically relating to a CUP-VISAR compressed image inversion method and device. Background Technology

[0002] Laser-driven inertial confinement fusion (ICF) research is a challenging field at the forefront of international science and one of the important methods for achieving controlled nuclear fusion. Currently, the asymmetry of implosion compression is considered a significant cause of ICF ignition failure. The ICF shock wave velocity can be used to calculate entropy increase and predict the compression state of the target pellet, among other important information.

[0003] Imaging velocity interferometer systems for any reflector (VISAR) are widely used in the diagnosis of ICF shock waves, but they can only acquire one-dimensional fringe information of the shock wave front. In 2020, Yang et al. proposed a CUP-VISAR diagnostic system that combines compressed ultrafast photography (CUP) with VISAR. Through compressed sensing (CS) technology, they were able to reconstruct multiple high-temporal-resolution time-varying two-dimensional fringe images from a single compressed image, thus promoting the development of high spatiotemporal resolution two-dimensional implosion diagnostic technology.

[0004] In 2021, Guan et al. proposed a CUP-VISAR data simulation method to study the length of the equipment diagnostic time window. First, a series of original VIASAR images were generated, and then the compressed images were simulated. Finally, a two-step iterative shrinkage thresholding algorithm was used to reconstruct the compressed images, obtaining the corresponding reconstructed images. Then, the velocity error and image correlation coefficient values ​​were calculated, providing a reference for actual experiments.

[0005] Currently, in the CUP-VISAR field, the two-step iterative threshold algorithm (TwIST) based on total variation (TV) regularization is the mainstream algorithm for shock wave stripe image inversion, but it has the following drawbacks: 1. When there are a large number of compressed shock wave stripe images, a lot of detailed information in the inverted image is lost, resulting in over-smoothing of the image and a significant reduction in the spatial resolution of the image inverted by the algorithm.

[0006] 2. When the striped image is affected by large Gaussian noise, the quality of the inverted image is greatly reduced, resulting in unclear outlines of the striped image.

[0007] 3. The current algorithm is greatly affected by the regularization parameter, which can easily lead to unstable inversion results. Summary of the Invention

[0008] To address the aforementioned problems in the prior art, this invention provides a CUP-VISAR compressed image inversion method and device, which improves image quality.

[0009] The present invention adopts the following technical solution: A CUP-VISAR compressed image inversion method includes the following steps: Obtain a two-dimensional striped compressed image and represent the two-dimensional striped compressed image using mathematical equations; By adding low-rank prior constraints and total variation constraints, the mathematical equations are solved in reverse to invert the two-dimensional fringe compressed image and obtain the original interference fringe image. The feasibility of the inversion method was verified.

[0010] As a further improvement to the present invention, the mathematical equation is: (1) A two-dimensional striped compressed image. For spacetime integral operators, For time shearing operator, For spatial coding operators, It is the original interference fringe image at the imaging point.

[0011] As a further improvement of the present invention, before the step of solving the mathematical equation in reverse, the following step is also included: Add prior constraints to the original interference fringe image to be determined, transforming it into an objective function. Solving the unconstrained optimization problem: (2) in, express Norm, For regularization functions, This is the regularization parameter.

[0012] As a further improvement of the present invention, when the objective function When adding low-rank prior constraints and total variational constraints, formula (2) is expressed in the following form: (3) in, For regularization parameters, For total variational regularization functions, This represents the rank function.

[0013] As a further improvement of the present invention, Convex relaxation is the nuclear norm, expressed in the following form as formula (3): (4) in, represents the nuclear norm of a matrix, and represents the sum of all singular values ​​of a matrix.

[0014] As a further improvement to the present invention, a set of auxiliary variables is added. By splitting variables, let , , Formula (4) is expressed in the following form: (5) in, Represents the dual variable. This represents the penalty parameter.

[0015] As a further improvement of the present invention, the step of inverting the mathematical equation to invert the two-dimensional fringe compressed image and obtain the original interference fringe image includes: Update the original variable

[0016] (6) Update using gradient descent. Optimize subproblems; (7) in, It is the identity matrix. For TV noise reduction operators; For the variables in formula (6) The update applies the singular value thresholding algorithm to the optimization problem of minimizing the nuclear norm, resulting in: (8) in, The soft threshold function is defined as follows: ,and Represents a symbolic function; , , yes The singular value decomposition result, i.e. ; Update 3D dynamic scene : (9) Solving equation (9) using the gradient descent method, we obtain: (10) Update penalty parameters : (11) in, , representing the original residual. , representing the dual residual. Represents the balance factor. This indicates the residual tolerance. It is typically taken as... , ; Check whether the iteration satisfies the convergence condition; If the iteration satisfies ,in The preset error value, and If the iteration satisfies the termination condition; If the iteration does not meet the convergence condition, update the dual variable. The final reconstruction result is obtained when the iteration meets the convergence condition: (12) As a further improvement to the present invention, the feasibility of the inversion method is verified by using a shooting experiment or a simulation experiment.

[0017] Furthermore, the present invention also provides a CUP-VISAR compressed image inversion device, comprising: a CUP-VISAR system and a computer device, wherein the computer device includes a processor and a memory, and the memory stores at least one instruction, at least one program, code set or instruction set, wherein the at least one instruction, at least one program, code set or instruction set is loaded and executed by the processor to implement the above-described CUP-VISAR compressed image inversion method.

[0018] As a further improvement of the present invention, the present invention also includes: an experimental device connected to the CUP-VISAR system.

[0019] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention recovers the low-rank matrix by adding low-rank prior constraints and total variational constraints, transforming the constraint into minimizing the sparsity of singular values ​​of the nuclear norm minimization matrix. The rank of the matrix is ​​the number of vectors in the maximal column vector-independent set of the matrix. If the rank of the matrix is ​​much smaller than its size, the matrix is ​​low-rank. Shock wave stripe images contain a large amount of redundant information and have similar spatial structures, which means that the stripe images are low-rank or approximately low-rank. Moreover, stripe images possess low-rank characteristics, while noise does not. Applying low-rank constraints to the stripe image matrix can achieve noise reduction while inverting the stripe image, thereby improving image quality. Attached Figure Description

[0020] The technology of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments: Figure 1 This is a flowchart of the CUP-VISAR compressed image inversion method described in Example 1; Figure 2 This is a flowchart of step S2 as described in Example 1; Figure 3 This refers to the original stripe image data in Example 1; Figure 4 The inversion result is obtained using the CUP-VISAR compressed image inversion method described in Example 1; Figure 5 The inversion results are obtained using the TV-TwIST algorithm; Figure 6 The PSNR curve of the reconstructed image in Example 1; Figure 7 The SSIM curve of the reconstructed image in Example 1; Figure 8 The encoded mask image in Example 1; Figure 9 The image is a one-dimensional stripe image recorded by the stripe camera of the experimental setup in Example 1; Figure 10 The image is a dynamic stripe image recorded by an external CCD camera in the CUP system of Example 1. Figure 11 To reconstruct images using the TV-TwIST algorithm; Figure 12 The image is an experimentally reconstructed image obtained using the CUP-VISAR compressed image inversion method described in Example 1; Figure 13 The VISAR image is extracted from the experimentally reconstructed image obtained using the TV-TwIST algorithm; Figure 14The VISAR image is extracted from the experimental reconstructed image obtained using the CUP-VISAR compressed image inversion method described in Example 1; Figure 15 The shock wave velocity curve and relative error diagram are shown in Example 1. Figure 16 This is a schematic diagram of the CUP-VISAR compressed image inversion device described in Example 2; Figure 17 This is another schematic diagram of the CUP-VISAR compressed image inversion device described in Example 2.

[0021] Figure label: 1. Optical path equipment; 2. Interferometer; 3. CUP system; 31. Streak camera; 4. Experimental setup; 41. Streak camera; 5. Probe light source; 6. Target point. Detailed Implementation

[0022] The following will provide a clear and complete description of the concept, specific structure, and technical effects of the present invention in conjunction with embodiments and accompanying drawings, so as to fully understand the purpose, solution, and effects of the present invention. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The same reference numerals used throughout the accompanying drawings indicate the same or similar parts.

[0023] It should be noted that, unless otherwise specified, when a feature is referred to as "fixed" or "connected" to another feature, it can be directly fixed or connected to the other feature, or indirectly fixed or connected to the other feature. Furthermore, the descriptions of "up," "down," "left," and "right" used in this invention are only relative to the relative positional relationships of the various components of the invention in the accompanying drawings.

[0024] Example 1 This embodiment provides a CUP-VISAR compressed image inversion method, such as Figure 1 As shown, the steps include: S1. Obtain the two-dimensional stripe compressed image and represent it using mathematical equations: (1) A two-dimensional striped compressed image. For spacetime integral operators, For time shearing operator, For spatial coding operators, It is the original interference fringe image at the imaging point.

[0025] Considering that the number of equations in formula (1) is much smaller than the number of unknowns, it belongs to the problem of solving an underdetermined problem. Therefore, by adding prior constraints to the original interference fringe image to be solved, it is transformed into an objective function. Solving the unconstrained optimization problem: (2) in, express Norm, For regularization functions, This is the regularization parameter.

[0026] When the objective function When adding low-rank prior constraints and total variational constraints, formula (2) is expressed in the following form: (3) in, For regularization parameters, For total variational regularization functions, This represents the rank function.

[0027] Due to the rank function The process is an NP-hard problem and cannot be solved concretely, therefore... Convex relaxation is the nuclear norm, expressed in the following form as formula (3): (4) in, Let represent the nuclear norm of the matrix, and let represent the sum of all singular values ​​of the matrix.

[0028] The PnP-ADMM iterative algorithm can split the objective function by means of variable splitting. Decomposing the problem into several sub-problems allows for a more efficient solution to the proposed optimization problem. Adding a set of auxiliary variables... By splitting variables, let , , Formula (4) is expressed in the following form: (5) in, Represents the dual variable. This represents the penalty parameter.

[0029] In step S1, a CUP-VISAR compressed image inversion model based on the PnP-ADMM solution framework is established. Simultaneously, a low-rank prior constraint is added to the shock wave fringe image, and the low-rank matrix is ​​recovered by minimizing the sparsity of the singular values ​​of the constraint matrix using the nuclear norm minimization constraint.

[0030] S2. Add low-rank prior constraints and total variation constraints, and solve the mathematical equations in reverse to invert the two-dimensional fringe compressed image, obtaining the original interference fringe image. Figure 2 As shown, it specifically includes: S21. Update the original variables.

[0031] (6) Update using gradient descent Optimize subproblems; (7) in, It is the identity matrix. For TV noise reduction operators; For the variables in formula (6) The update applies the singular value thresholding algorithm to the optimization problem of minimizing the nuclear norm, resulting in: (8) in, The soft threshold function is defined as follows: ,and Represents a symbolic function; , , yes The singular value decomposition result, i.e. ; S22, Update the 3D dynamic scene : (9) Solving equation (9) using the gradient descent method, we obtain: (10) S23, Update penalty parameters : (11) in, , representing the original residual. , representing the dual residual. Represents the balance factor. This indicates the residual tolerance. It is typically taken as... , ; S24. Check if the iteration meets the convergence condition, such as Figure 2 As shown: If the iteration satisfies ,in The preset error value, and If the iteration satisfies the termination condition; If the iteration does not meet the convergence condition, update the dual variable. The final reconstruction result is obtained when the iteration meets the convergence condition: (12) S3. Verify the feasibility of the inversion method.

[0032] Experiment 1 This embodiment utilizes MATLAB to conduct simulation experiments on the proposed inversion method, analyzing and comparing the reconstruction performance of the mainstream algorithm TV-TwIST. Since the temporal resolution of the CUP-VISAR diagnostic system is determined by the streak camera, the main focus is on verifying the proposed algorithm's ability to invert two-dimensional images. The simulation simulates the streak image compression process, including encoding the original interferometric fringe image using DMD, spatiotemporal shearing by the streak camera, and CCD superposition imaging, followed by image inversion using the proposed algorithm. Peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) are selected as evaluation metrics for the quality of the inverted image.

[0033] Simulated original interference fringe image as Figure 3 As shown, the size is The interference fringe pattern data contains rich information along the time dimension. As you can see, the shock wave went through stages of rest, deceleration, and acceleration, which is highly representative. To more closely resemble a real, complex environment, the DMD simulation employs... The encoding matrix for the encoded aperture is randomly sampled from the original stripe image. Simultaneously, Gaussian noise with a mean of 0 and a variance of 0.1 is introduced into the stripe image data.

[0034] The CUP-VISAR system's data acquisition process is simulated by spatial coding, temporal cropping, and spatiotemporal compression of the original fringe image to obtain a noisy observation image. The original fringe image is then inverted using the TV-TwIST algorithm and the proposed method. The inversion result is shown below. Figure 4 and Figure 5 As shown. From a subjective visual perspective, both algorithms can reconstruct the image, but the image reconstructed by the TV-TwIST algorithm is relatively blurry, and it is difficult to reconstruct the stripe details in areas where the shock wave changes significantly (around frame 9). The proposed algorithm has a clearer overall visual effect, and its reconstruction of the stripe contour details is also superior. To quantitatively represent the quality of the reconstructed image, Figure 6 , Figure 7The PSNR and SSIM of the reconstructed images are presented. The PSNR of the images reconstructed by the TV-TwIST algorithm ranges from 8.61 dB to 18.9 dB, and the SSIM ranges from 57.6% to 82.6%. The proposed algorithm reconstructs images with PSNR ranging from 14.8 dB to 27.6 dB and SSIM ranging from 70.9% to 89.2%. The average PSNR is improved by 8.45 dB, and the average SSIM is improved by 8.52%. Simulation results show that the proposed reconstruction algorithm has superior reconstruction quality.

[0035] As shown in Experiment 1, compared with the existing TV-TwIST method, the CUP-VISAR compressed image inversion method of this embodiment has significantly improved the image quality and stronger noise resistance in the CUP-VISAR diagnostic system.

[0036] Experiment 2 The feasibility of the inversion method was verified using a target shooting experiment. Due to limitations in the experiment, the DMD used a coding aperture of [missing information - likely a specific aperture value]. The pixel's encoding mask aperture also affects the imaging due to speckle noise. Therefore, considering the complexity of actual experiments, the number of fringes was chosen to be 3.

[0037] In the experiment, the encoding aperture was... Pixel-encoded mask image as Figure 8 As shown, the pixel size is Images recorded by the streak camera, such as... Figure 9 , Figure 10 As shown, (a) is a one-dimensional fringe image recorded by a fringe camera with added branches, and (b) is a dynamic fringe image recorded by a CCD camera external to the fringe camera in the CUP system, with a pixel size of [missing information]. In the experiment, the scanning mode of the streak camera was 200 ns, and the pulse width of the probe light was 5 ns, so the number of reconstructed images was 25 frames.

[0038] The dynamic experimental images were reconstructed using both the TV-TwIST algorithm and the proposed algorithm. The reconstruction results are as follows: Figure 11 , Figure 12 As shown, each row of the reconstructed image is extracted and arranged in chronological order to reconstruct a one-dimensional VISAR image. The one-dimensional VISAR image shown is obtained by extracting the middle row and reconstructing it. Figure 13 and Figure 14 The shock wave velocities extracted from the VISAR images shown are respectively in Figure 15 The text indicates that... Figure 13 The shock wave velocity extracted from the one-dimensional VISAR image is used as a reconstruction error reference to obtain the relative error of the shock wave velocity. The relative error curve is shown in... Figure 15The maximum relative error obtained by the TV-TwIST algorithm is less than 13.5%, while the maximum relative error of the proposed algorithm is less than 3.46%, a reduction of 10.04%. Experimental results show that the CUP-VISAR compressed image inversion method of this embodiment can restore the dynamic scene of the CUP-VISAR diagnostic system and obtain higher-precision shock wave velocity data, which has practical application value.

[0039] As shown in Experiment 2, the relative velocity error of the shock wave extracted by the CUP-VISAR compressed image inversion method in this embodiment is significantly reduced, and higher precision shock wave velocity data can be obtained, which has practical application value.

[0040] Example 2 This embodiment provides a CUP-VISAR compressed image inversion device, such as... Figure 16 The system includes a CUP-VISAR system and a computer device. The computer device includes a processor and a memory. The memory stores at least one instruction, at least one program, code set, or instruction set. The at least one instruction, at least one program, code set, or instruction set is loaded and executed by the processor to implement the CUP-VISAR compressed image inversion method as described in Embodiment 1.

[0041] The CUP-VISAR system comprises, in sequence, optical path equipment, an interferometer, and a CUP system. For example... Figure 3 As shown, the solid arrows indicate the optical path of the probe light in the system. First, the probe light emitted from the probe source passes sequentially through lens L1, compensation lens COM1, and beam splitter BS1 to reach reflector M1 and is reflected. Then, the reflected light is reflected by reflector M2 and focused onto the target point by lenses L2 and L3. Subsequently, the probe light carrying Doppler frequency shift information returns along the original path after reflection at the target point, passing sequentially through lens L3, lens L2, reflector M2, reflector M1, beam splitter BS1, reflector M3, and lens L4, before entering the interferometer composed of beam splitter BS2, reflector M5, reflector M4, beam splitter BS3, reflector M6, and a standard for time-delay mixing. The mixed interference light is focused and imaged at imaging point IP2 by lens L5 and enters the CUP system. Imaging point IP2 is imaged by the first 4f imaging system. The image is projected onto a digital micromirror device (DMD) and combined to generate an coded image. Next, the coded image is projected onto the slit of a streak camera via a second set of 4f optical paths. The slit of the streak camera is fully open (~5 mm), and two-dimensional streak images at different times are scanned along the time direction to record the two-dimensional information of the object point. Then, the offset coded image is spatiotemporally compressed on an external CCD camera of the streak camera to obtain a compressed two-dimensional streak image.

[0042] like Figure 17 The embodiment further includes an experimental setup connected to the CUP-VISAR system. The slit of the stripe camera on the experimental setup is not fully open (approximately 100 μm), and the recorded one-dimensional VISAR image is used as a velocity error reference for the shock wave.

[0043] The processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0044] The memory can be used to store the computer programs or modules. The processor implements various functions of the assistive terminal device based on mirror neuron therapy by running or executing the computer programs or modules stored in the memory and calling data stored in the memory. The memory can mainly include a program storage area and a data storage area. The program storage area can store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory can include high-speed random access memory, and can also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0045] As can be seen from the above embodiments, the present invention has the following beneficial effects: 1. The plug-and-play alternate direction multiplier method (PnP-ADMM) can decompose the target inversion problem into several subproblems through variable splitting, enabling faster and more accurate solutions. Furthermore, due to its plug-and-play nature, it is easy to add more prior constraints to the inversion problem, thereby improving the accuracy of the inverted image.

[0046] 2. The rank of a matrix is ​​the number of vectors in the maximal column vector-independent set of the matrix. If the rank of a matrix is ​​much smaller than its size, the matrix is ​​low-rank. Shock wave stripe images contain a large amount of redundant information and have similar spatial structures, which means that the stripe images are low-rank or approximately low-rank. Moreover, stripe images have low-rank characteristics, while noise does not. Therefore, imposing a low-rank constraint on the stripe image matrix can achieve noise reduction while inverting the stripe image.

[0047] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Therefore, any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A CUP-VISAR compressed image inversion method, characterized in that, Including the following steps: Obtain a two-dimensional striped compressed image and represent the two-dimensional striped compressed image using mathematical equations; By adding low-rank prior constraints and total variation constraints, the mathematical equations are solved in reverse to invert the two-dimensional fringe compressed image and obtain the original interference fringe image. The feasibility of the inversion method is verified; The mathematical equation is: (1) A two-dimensional striped compressed image. For spacetime integral operators, For time shearing operator, For spatial coding operators, It is the original interference fringe image at the imaging point; Before the step of solving the mathematical equation in reverse, the following step is also included: Add prior constraints to the original interference fringe image to be determined, transforming it into an objective function. Solving the unconstrained optimization problem: (2) in, express Norm, For regularization functions, For regularization parameters; When the objective function When adding low-rank prior constraints and total variational constraints, formula (2) is expressed in the following form: (3) in, For regularization parameters, For total variational regularization functions, Represents the rank function; Will Convex relaxation is the nuclear norm, expressed in the following form as formula (3): (4) in, The nuclear norm of a matrix is ​​denoted by , and the sum of all singular values ​​of the matrix is ​​denoted by . Add a set of auxiliary variables By splitting variables, let , , Formula (4) is expressed in the following form: (5) in, Represents the dual variable. Indicates the penalty parameter; The step of inverting the mathematical equation to obtain the original interference fringe image by reversing the two-dimensional fringe compressed image includes: Update the original variable ; (6) Update using gradient descent Optimize subproblems; (7) in, It is the identity matrix. For TV noise reduction operators; For the variables in formula (6) The update applies the singular value thresholding algorithm to the optimization problem of minimizing the nuclear norm, resulting in: (8) in, The soft threshold function is defined as follows: ,and Represents a symbolic function; , , yes The singular value decomposition result, i.e. ; Update 3D dynamic scene : (9) Solving equation (9) using the gradient descent method, we obtain: (10) Update penalty parameters : (11) in, , representing the original residual, , representing the dual residual, Represents the balance factor. , Indicates the residual tolerance. ; Check whether the iteration satisfies the convergence condition; If the iteration satisfies ,in The preset error value, ,and , If the iteration satisfies the convergence condition; If the iteration does not meet the convergence condition, update the dual variable. The final reconstruction result is obtained when the iteration meets the convergence condition: (12)。 2. The CUP-VISAR compressed image inversion method according to claim 1, characterized in that, The feasibility of the inversion method is verified by using shooting experiments or simulation experiments.

3. A CUP-VISAR compressed image inversion device, characterized in that, Includes: a CUP-VISAR system and a computer device, the computer device including a processor and a memory, the memory storing at least one instruction, at least one program, code set or instruction set, the at least one instruction, at least one program, code set or instruction set being loaded and executed by the processor to implement the CUP-VISAR compressed image inversion method as described in claim 1 or 2.

4. The CUP-VISAR compressed image inversion device according to claim 3, characterized in that, Also includes: Experimental setup connected to the CUP-VISAR system.

Images