Compression imaging method and system based on point spread function and dislocation superposition

Abstract

The invention discloses a point spread function and dislocation superposition-based compression imaging method and system, and relates to the technical field of compression imaging, and the method comprises the specific steps: carrying out the PSF imaging of a collected image based on a point spread function, and obtaining a blurred image; encoding the blurred image based on an encoding matrix to obtain an encoded image; carrying out frame-by-frame dislocation superposition on the coded image to obtain an original image; and constructing a measurement matrix based on the original image, and decoding and reconstructing the original image by using a reconstruction algorithm according to the measurement matrix to obtain a reconstructed image. According to the method, a high-speed compression imaging model based on PSF diffusion, frosted glass coding and frame-by-frame staggered superposition is constructed on the basis of a compressed sensing theory, and multiple frames of images are superposed on a photosensitive chip of a camera in a staggered manner by the model between two electronic reading times of the camera. Therefore, the electronic reading speed of the camera is improved in a disguised manner.

Description

Technical Field

[0001] The present invention relates to the technical field of compression imaging, and more particularly to a compression imaging method and system based on point spread function and staggered superposition. Background Art

[0002] With the development of science and technology, the demand for high-speed imaging is increasing. In many fields such as remote sensing photography, traffic monitoring, mechanical analysis, and life sciences, capturing rapidly changing events requires the use of cameras with high temporal resolution. However, traditional high-speed cameras face some technical challenges:

[0003] Temporal resolution limitation: The temporal resolution of a high-speed camera is limited by its electronic readout speed. If you want to increase the temporal resolution, you have to increase the transfer rate of image data from the sensor to the processing unit, which is usually limited by the hardware performance.

[0004] Limited storage capacity: The data acquisition process of high-speed cameras relies on the support of storage devices such as memory and hard disk. Once the storage space is exhausted, it is impossible to continue recording subsequent images, which poses a serious obstacle to long-term continuous shooting tasks.

[0005] Point spread function (PSF) effect: In the actual imaging process, due to the influence of diffraction, scattering and other factors, the ideal point light source will form an extended distribution rather than a single point on the image plane after passing through the optical system. This phenomenon is called the PSF effect. The existence of PSF will cause image blur and reduce the quality and resolution of the final image. Therefore, how to use existing hardware resources to achieve longer shooting time and higher spatiotemporal resolution without sacrificing image quality is an urgent problem that technicians in this field need to solve. Summary of the invention

[0006] In view of this, the present invention provides a compression imaging method and system based on point spread function and staggered superposition, which overcomes the above-mentioned defects.

[0007] In order to achieve the above object, the present invention adopts the following technical solution:

[0008] A compression imaging method based on point spread function and offset superposition, the specific steps are:

[0009] Perform PSF imaging on the collected image based on the point spread function to obtain a blurred image;

[0010] Encoding the blurred image based on a coding matrix to obtain a coded image;

[0011] Performing frame-by-frame staggered superposition on the encoded images to obtain an original image;

[0012] A measurement matrix is ​​constructed based on the original image, and the original image is decoded and reconstructed using a reconstruction algorithm according to the measurement matrix to obtain a reconstructed image.

[0013] Optionally, the step of acquiring the coding matrix is: digitizing the frosted glass to obtain the coding matrix.

[0014] Optionally, the steps of constructing the measurement matrix are:

[0015] Constructing a block diagonal matrix using the coding measurement matrix as a sub-block matrix;

[0016] Merging data on columns within a specified region of the block diagonal matrix to obtain a first matrix;

[0017] The rows in the designated area of ​​the first matrix are arranged in a preset order, and data of repeated rows are merged to obtain the measurement matrix.

[0018] Optionally, the step of acquiring the encoding measurement matrix is:

[0019] Each column element in the point diffusion matrix is ​​multiplied pixel by pixel by the encoding matrix to obtain the encoding measurement matrix.

[0020] A compression imaging system based on point spread function and offset superposition, comprising:

[0021] An image blur module is used to perform PSF imaging on the collected image according to the point spread function to obtain a blurred image;

[0022] An image encoding module, used for encoding the blurred image according to a coding matrix to obtain a coded image;

[0023] An image superposition module, used for performing staggered superposition of the encoded image frame by frame to obtain an original image;

[0024] The image reconstruction module is used to construct a measurement matrix according to the original image, and decode and reconstruct the original image using a reconstruction algorithm according to the measurement matrix to obtain a reconstructed image.

[0025] Optionally, the image encoding module includes a digitization unit for digitizing the frosted glass to obtain the encoding matrix.

[0026] Optionally, the image reconstruction module includes:

[0027] A matrix construction unit, used for constructing a block diagonal matrix using a coding measurement matrix as a sub-block matrix;

[0028] A first processing unit is used to merge data of columns in a specified area of ​​the block diagonal matrix to obtain a first matrix;

[0029] A second processing unit, used for arranging the rows in the specified area of ​​the first matrix according to a preset order and merging the data of the repeated rows to obtain the measurement matrix;

[0030] The image decoding unit is used to decode and reconstruct the original image using a reconstruction algorithm according to the measurement matrix to obtain a reconstructed image.

[0031] Optionally, the matrix construction unit includes a coding measurement matrix construction subunit, which is used to multiply each column element in the point diffusion matrix by the coding matrix pixel by pixel to obtain the coding measurement matrix.

[0032] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a compression imaging method and system based on point spread function and staggered superposition. Based on the theory of compressed sensing, a high-speed compression imaging model based on PSF diffusion, ground glass coding and frame-by-frame staggered superposition is constructed. Between the two electronic reading times of the camera, the model staggers and superimposes multiple frames of images on the photosensitive chip of the camera. Thereby, the electronic reading speed of the camera is improved in disguise. At the same time, the amount of data acquisition is effectively compressed, which is equivalent to increasing the capacity of the storage device and the signal bandwidth. Thereby, the shooting time and time resolution of the high-speed camera are improved. Based on the compressed sensing reconstruction algorithm, the model efficiently reconstructs the blurred image of the staggered superposition with obvious PSF effect, thereby obtaining a clear reconstructed image. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0034] Figure 1 A schematic diagram of the method provided by the present invention;

[0035] Figure 2(a) shows the multi-frame images before encoding; Figure 2(b) shows the Bessel PSF; Figure 2(c) shows the frosted glass; Figure 2(d) shows the streak camera; Figure 2(e) shows the original image;

[0036] Figure 3 A schematic diagram of a coding matrix provided by the present invention;

[0037] FIG4(a) is a schematic diagram of a coding measurement matrix when the input data is one frame of image; FIG4(b) is a schematic diagram of a coding measurement matrix when the input data is three frames of image;

[0038] Figure 5(a) shows three blurred images based on PSF diffusion blurring when the half-maximum width of the PSF is 3.125 pixels; Figure 5(b) shows a blurred offset superposition image; Figure 5(c) shows a reconstructed image before left optimization obtained by reconstructing the blurred offset superposition image connected end to end in columns as measurement data through a reconstruction algorithm; Figure 5(d) shows a reconstructed image after left optimization obtained by reconstructing the blurred offset superposition image connected end to end in columns as measurement data through a reconstruction algorithm; Figure 5(e) shows an encoded image; Figure 5(f) shows an original image; Figure 5(g) shows a reconstructed image obtained by reconstructing the original image connected end to end in columns as measurement data through a reconstruction algorithm; Figure 5(h) shows a reconstructed image after left optimization obtained by reconstructing the original image connected end to end in columns as measurement data through a reconstruction algorithm;

[0039] Figure 6(a) shows three blurred images based on PSF diffusion blur when the half-maximum full width of the PSF is 0.7812 pixels; Figure 6(b) shows a blurred offset superposition image; Figure 6(c) shows a reconstructed image before left optimization obtained by reconstructing the blurred offset superposition image after connecting the end to the end of each column as the measurement data through the reconstruction algorithm; Figure 6(d) shows a reconstructed image after left optimization obtained by reconstructing the blurred offset superposition image after connecting the end to the end of each column as the measurement data through the reconstruction algorithm; Figure 6(e) shows an encoded image; Figure 6(f) shows an original image; Figure 6(g) shows a reconstructed image obtained by reconstructing the original image after connecting the end to the end of each column as the measurement data through the reconstruction algorithm; Figure 6(h) shows a reconstructed image after left optimization obtained by reconstructing the original image after connecting the end to the end of each column as the measurement data through the reconstruction algorithm. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] This embodiment discloses a compression imaging method based on point spread function and offset superposition, such as Figure 1 As shown, the specific steps are:

[0042] Step 1, performing PSF imaging on multiple frame images based on the point spread function to obtain a blurred image;

[0043] Step 2: Encode the blurred image based on the encoding matrix to obtain an encoded image;

[0044] Step 3, performing staggered superposition of the encoded images frame by frame to obtain the original image;

[0045] Step 4: Construct a measurement matrix based on the original image, and decode and reconstruct the original image using a reconstruction algorithm according to the measurement matrix to obtain a reconstructed image.

[0046] In one embodiment, the step of obtaining the encoding matrix is: digitize the frosted glass to obtain the encoding matrix.

[0047] Further, a compressive imaging method based on the point spread function and misaligned superposition is implemented based on a high-speed compressive imaging model, and its structure includes Figure 2(b)-Figure 2(d) , where the input data is shown in Fig. 2(a), which is multiple frames of images before encoding, that is, the true image, with an image size of 32×32 pixels; Fig. 2(b) is the Bessel PSF (full width at half maximum is 3.125 pixels); Fig. 2(c) is the frosted glass; the matrix after digitizing the frosted glass is as Figure 3 shown. The frosted glass is a random mask. The structured light generated through the frosted glass can encode the image; Fig. 2(d) is a streak camera; the output data is shown in Fig. 2(e), which is the original image after frame-by-frame misaligned superposition, and the misalignment step is 8 pixels. The streak camera converts the time distribution of the optical signal into a spatial distribution by shearing photoelectrons, and can achieve an extremely high frame rate. In this model, the streak camera is used to implement the misaligned superposition function. The streak camera can also be replaced by other high-speed cameras, and the equivalent misaligned superposition function can be implemented by software programming for the encoded multiple frames of images collected by the camera.

[0048] Furthermore, the ratio of the number of pixels of the 3 frames of images in Fig. 2(a) to the number of pixels of the original image in Fig. 2(e) is the compression ratio. The image compression ratio corresponding to the misalignment step of 8 pixels is 2, and the corresponding shooting duration is increased by two times. Between the two electron reading times of the camera, only one frame of image could be collected in the past, but now three frames of images can be collected. Therefore, the time resolution is increased by two times.

[0049] Among them, the mathematical model expression of compressive sensing is:

[0050] min||x||0 s.t. y = Ax (1);

[0051] In the formula: y is the measurement data, y ∈ R M ; x is the signal, x ∈ R N ; after binarizing x, the output data X is obtained; A is the measurement matrix, M < N; min||x||0 is the objective function; y = Ax is the constraint function; ||·||0 is the l0 norm, and the objective function can also use the l1 or l TV norm.

[0052] In one embodiment, the step of constructing the measurement matrix is:

[0053] Step 41: Construct a block diagonal matrix with the encoding measurement matrix as the sub-block matrix;

[0054] Step 42: merge the columns in the specified area of ​​the block diagonal matrix to obtain a first matrix;

[0055] Step 43: Arrange the rows in the designated area of ​​the first matrix according to a preset order, and merge the data of the repeated rows to obtain a measurement matrix.

[0056] In one embodiment, the steps of obtaining the encoding measurement matrix are:

[0057] Each column element in the point diffusion matrix is ​​multiplied pixel by pixel with the encoding matrix to obtain the encoding measurement matrix.

[0058] Furthermore, in formula (1), the one-dimensional vectors y and x are respectively composed of the original image and the true image connected column by column. When the input data has only one frame of image and there is no frosted glass element in the optical path, the measurement matrix is ​​determined by the PSF of the imaging system. The jth column of the PSF-based measurement matrix P corresponds to the original image captured by the camera when only the position index j of x emits light. G is a vector composed of the matrix based on frosted glass connected column by column. When there is a frosted glass element in the optical path, the matrix obtained by multiplying each column of the measurement matrix by G pixel by pixel is the coded measurement matrix P' when there is a frosted glass element, as shown in Figure 4(a).

[0059] When the input data has three frames of images, the three frames of original images collected by the camera are offset and superimposed. After the original image and the true image after offset superposition are connected column by column into vectors, the corresponding measurement matrix and generation method are shown in Figure 4(b). First, the coding measurement matrix P' is used as the sub-block matrix to construct a three-block block diagonal matrix; secondly, by shifting each column, the columns of the coding measurement matrix P' correspond to the positions of the elements of x; finally, by shifting each row, the rows of the coding measurement matrix P' correspond to the positions of the elements of y.

[0060] Furthermore, various data are obtained, including: a vector composed of the matrix based on frosted glass connected column by column, namely the encoding matrix G; a measurement matrix based on PSF, namely the point spread matrix P, the staggered step size S, and the number of image frames L.

[0061] The steps to construct the block diagonal matrix PP are:

[0062] Get the matrix size of the point diffusion matrix P, specifically: the point diffusion matrix P is an M×N matrix, where M is the number of rows and N is the number of columns; len_y and len_x are the square roots of M and N respectively;

[0063] Multiply each column element in the point diffusion matrix P by the encoding matrix G pixel by pixel to obtain the encoding measurement matrix P';

[0064] A block diagonal matrix PP is constructed with the coding measurement matrix P' as a sub-block matrix, wherein the number of the sub-block matrices is L.

[0065] In step 42, the first matrix B is generated by:

[0066] Iterative execution merges the ((i-1)×len_x+1+len_x×len_x×(j-1))th to (i×len_x+len_x×len_x×(j-1))th columns of the block diagonal matrix PP from left to right into a new matrix, i.e., the first matrix B; where i is the number of outer iterations (traversing rows) and j is the number of inner iterations (traversing column blocks);

[0067] In step 42, the steps of generating the measurement matrix A are:

[0068] Iterative execution arranges the rows ((i-1)×len_y+1+len_y×len_y×(j-1)) to (i×len_y+len_y×len_y×(j-1)) of the first matrix B in the order of step×(j-1)+1+(i-1)×MM to step×(j-1)+len_y+(i-1)×MM, and the repeated rows are accumulated to obtain the measurement matrix A. Among them, i is the number of outer iterations (traversing rows) and j is the number of inner iterations (traversing column blocks).

[0069] MM is the total number of rows of the superimposed image, taking into account the offset step S and the number of frames L. The specific expression is: MM = len_y + S × (L-1).

[0070] In one embodiment, the original image after connecting the columns end to end is used as the measurement data in formula (1), and based on the measurement matrix A, the TVAL3 algorithm (Total Variation Minimization by Augmented Lagrangian and Alternating Direction Algorithms) of compressed sensing is used for decoding and reconstruction, and the obtained clear image is shown in Figure 5(g). The SSIM and SNR reach 0.79 and 17.05 dB respectively, achieving a good reconstruction effect.

[0071] The left optimization algorithm of compressed sensing can achieve the purpose of improving the reconstruction effect by optimizing the measurement matrix and measurement data afterward without changing the hardware design and the measured data already collected. The optimized matrix can be obtained by performing row orthogonal normalization on the measurement matrix. Since the row orthogonal normalization operation belongs to the elementary transformation in matrix theory, the optimized matrix is ​​equivalent to the measurement matrix. Performing elementary row transformation on the measurement matrix is ​​equivalent to multiplying the measurement matrix on the left by the corresponding M×M elementary matrix (i.e., operator matrix T). The operator matrix T can be obtained by T=A OA T (AA T ) -1 The mathematical model of compressed sensing based on the left optimization algorithm is equivalently transformed from formula (1), and the expression is:

[0072] min||x|| TV st y T =A O x (2);

[0073] In the formula, y T =Ty,A O =TA,Ty=TAx.

[0074] Figure 5(h) is a clear image reconstructed by TVAL3 decoding after left optimization of Figure 5(f) and the measurement matrix. The SSIM and SNR reach 0.93 and 19.4 dB, respectively, which are 0.14 and 2.35 dB higher than those of Figure 5(g), achieving a good reconstruction effect. The left optimization algorithm can effectively improve the reconstruction effect.

[0075] In order to verify the influence of frosted glass on the performance of high-speed compression imaging model, in the optical path design, the frosted glass original as shown in Figure 2(c) is directly removed, and the frosted glass encoding link is abandoned. Figure 5(a) shows three blurred images after PSF diffusion blur; after the streak camera offsets and superimposes each frame of the image in Figure 5(a), a blurred offset superimposed image as shown in Figure 5(b) is obtained. Figures 5(c) and 5(d) are the clear images before and after the left optimization reconstructed by the TVAL3 algorithm, using the blurred offset superimposed images after connecting the end to the end of each column as the measurement data. Figure 5(c) shows the image before left optimization, with SSIM and SNR of 0.5 and 9.96dB respectively; Figure 5(d) shows the image after left optimization, with SSIM and SNR of 0.83 and 16.85dB respectively, which are improved by 0.33 and 6.89dB respectively. The left optimization algorithm significantly improves the reconstruction effect.

[0076] Comparing the reconstruction effects before and after adding the frosted glass element in the optical path (as shown in Figure 5(c) and Figure 5(g)), the SSIM and SNR of Figure 5(g) are increased by 0.29 and 7.09 dB respectively. Therefore, adding the frosted glass element in the optical path significantly improves the reconstruction effect. Comparing Figure 5(d) and Figure 5(h), the SSIM and SNR of Figure 5(h) are increased by 0.1 and 2.55 dB respectively. Frosted glass coding also has a significant positive improvement on the reconstruction effect after left optimization.

[0077] Figure 6(a)-Figure 6(h) The half-maximum width of the PSF used is 0.7812 pixels, which is much smaller than Figure 5(a)-Figure 5(h)The 3.125 pixels used in the image are 0.99 and 29.45 dB, respectively, which are almost identical to the true image, and no obvious PSF effect causing blur is found. As shown in Figures 6(c) and 6(g), the SSIM and SNR before and after frosted glass coding are increased by 0.16 and 4.99 dB, respectively. Frosted glass coding can effectively improve the reconstruction effect. As shown in Figures 6(c) and 6(d), the SSIM and SNR of the two are almost the same. The left optimization cannot improve the reconstruction effect when the PSF effect is not obvious, and Figures 6(g) and (h) are similar. Therefore, it is not difficult to see that the left optimization can improve the reconstruction effect when the PSF effect is obvious.

[0078] Comparing Figure 5(c) and Figure 6(c), it can be found that the SSIM and SNR of Figure 6(c) are 0.07 and 0.61dB higher, respectively. The obvious PSF effect does not directly improve the reconstruction effect. Comparing Figure 5(g) and Figure 6(g), it can be found that the SSIM and SNR of Figure 5(g) are increased by 0.06 and 1.49dB, respectively. The obvious PSF effect and frosted glass coding combination can achieve better reconstruction effect. Comparing Figure 5(h) and Figure 6(h), it can be found that the SSIM and SNR of Figure 5(h) are increased by 0.2 and 4.7dB, respectively. The obvious PSF effect, frosted glass and left optimization combination can achieve excellent reconstruction effect. The window structure in Figure 6(h) is unclear, while the window details in Figure 5(h) are clear and neat.

[0079] On the other hand, this embodiment further discloses a compression imaging system based on point spread function and offset superposition, including:

[0080] An image blur module is used to perform PSF imaging on multiple frames of images according to a point spread function to obtain a blurred image;

[0081] An image encoding module, used for encoding the blurred image according to the encoding matrix to obtain an encoded image;

[0082] An image overlay module is used to perform staggered overlay of the encoded images frame by frame to obtain the original image;

[0083] The image reconstruction module is used to construct a measurement matrix according to the original image, and decode and reconstruct the original image using a reconstruction algorithm according to the measurement matrix to obtain a reconstructed image.

[0084] In one embodiment, the image encoding module includes a digitizing unit for digitizing the frosted glass to obtain a coding matrix.

[0085] In one embodiment, the image reconstruction module includes:

[0086] A matrix construction unit, used for constructing a block diagonal matrix using a coding measurement matrix as a sub-block matrix;

[0087] A first processing unit is used to merge data of columns in a specified area of ​​the block diagonal matrix to obtain a first matrix;

[0088] A second processing unit, used for arranging the rows in the specified area of ​​the first matrix according to a preset order, and merging the data of the repeated rows to obtain a measurement matrix;

[0089] The image decoding unit is used to decode and reconstruct the original image using a reconstruction algorithm according to the measurement matrix to obtain a reconstructed image.

[0090] In one embodiment, the matrix construction unit includes a coding measurement matrix construction subunit, which is used to multiply each column element in the point diffusion matrix by the coding matrix pixel by pixel to obtain the coding measurement matrix.

[0091] In this embodiment, the PSF effect is prevalent in various images. When the PSF is large, the image quality is seriously affected, resulting in image blur. Based on the blurred image with obvious PSF effect, a high-speed compression imaging model and a measurement matrix construction method based on PSF, frosted glass and misalignment superposition are proposed. The obvious PSF effect will not only not degrade the reconstruction effect of the original image, but make the reconstructed image clearer. Experiments show that when the half-maximum full width of the PSF is 3.125 pixels and the compression ratio is 2, a reconstruction effect of SSIM and SNR of 0.79 and 17.05dB can be obtained. When the left optimization algorithm is used to perform left optimization processing on the original image and the measurement matrix, a reconstruction effect of SSIM and SNR of 0.93 and 19.4dB can be obtained. The high-speed compression imaging model can effectively compress data, extend shooting time and improve temporal resolution while ensuring the quality of reconstruction.

[0092] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0093] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A compression imaging method based on point spread function and offset superposition, characterized in that: The specific steps are: Perform PSF imaging on the collected image based on the point spread function to obtain a blurred image; Encoding the blurred image based on a coding matrix to obtain a coded image; Performing frame-by-frame staggered superposition on the encoded images to obtain an original image; A measurement matrix is ​​constructed based on the original image, and the original image is decoded and reconstructed using a reconstruction algorithm according to the measurement matrix to obtain a reconstructed image.

2. The method for compressive imaging based on point spread function and offset superposition according to claim 1, characterized in that: The step of acquiring the coding matrix is: digitizing the frosted glass to obtain the coding matrix.

3. The method for compressive imaging based on point spread function and offset superposition according to claim 1, characterized in that: The steps of constructing the measurement matrix are: Constructing a block diagonal matrix using the coding measurement matrix as a sub-block matrix; Merging data on columns within a specified region of the block diagonal matrix to obtain a first matrix; The rows in the designated area of ​​the first matrix are arranged in a preset order, and data of repeated rows are merged to obtain the measurement matrix.

4. The method for compressive imaging based on point spread function and offset superposition according to claim 3, characterized in that: The steps of obtaining the encoding measurement matrix are as follows: Each column element in the point diffusion matrix is ​​multiplied pixel by pixel by the encoding matrix to obtain the encoding measurement matrix.

5. A compression imaging system based on point spread function and offset superposition, characterized in that: include: An image blur module is used to perform PSF imaging on the collected image according to the point spread function to obtain a blurred image; An image encoding module, used for encoding the blurred image according to a coding matrix to obtain a coded image; An image superposition module, used for performing staggered superposition of the encoded image frame by frame to obtain an original image; The image reconstruction module is used to construct a measurement matrix according to the original image, and decode and reconstruct the original image using a reconstruction algorithm according to the measurement matrix to obtain a reconstructed image.

6. The compression imaging system based on point spread function and offset superposition according to claim 5, characterized in that: The image encoding module includes a digitizing unit for digitizing the frosted glass to obtain the encoding matrix.

7. The compression imaging system based on point spread function and offset superposition according to claim 5, characterized in that: The image reconstruction module comprises: A matrix construction unit, used for constructing a block diagonal matrix using a coding measurement matrix as a sub-block matrix; A first processing unit is used to merge data of columns in a specified area of ​​the block diagonal matrix to obtain a first matrix; A second processing unit, used for arranging the rows in the specified area of ​​the first matrix according to a preset order, and merging the data of the repeated rows to obtain the measurement matrix; The image decoding unit is used to decode and reconstruct the original image using a reconstruction algorithm according to the measurement matrix to obtain a reconstructed image.

8. The compression imaging system based on point spread function and offset superposition according to claim 7, characterized in that: The matrix construction unit comprises a coding measurement matrix construction subunit, which is used for multiplying each column element in the point diffusion matrix by the coding matrix pixel by pixel to obtain the coding measurement matrix.

Images