Image processing method and system based on deconvolution algorithm, and storage medium

Abstract

The invention discloses an image processing method and system based on a deconvolution algorithm, and a storage medium. The method comprises the steps of obtaining image data of a blurred image; generating a plurality of matched point spread functions through an optical microscopic system for capturing a blurred image; selecting an optimal point spread function and a preset number of iterations from the plurality of point spread functions; preprocessing the selected optimal point spread function; fourier transform is carried out on the point spread function through butterfly operation, and Fourier transform is carried out on the blurred image; performing deconvolution operation by using an iterative solution mode to obtain a target image; judging whether the number of iterations is reached or not; if yes, outputting a final target image; and if not, continuing repeated iteration until the number of iterations is reached, and outputting a final target image. According to the method, the limitation degree of the image is not high, the image information is not easy to lose, the requirement on equipment performance is low, the algorithm is optimized, the operation efficiency is improved, and the burden of equipment and resources is reduced.

Description

Technical Field

[0001] The present application relates to the technical field of image processing, and in particular, to an image processing method, system and storage medium based on a deconvolution algorithm. Background Art

[0002] In recent years, with the rapid development of optical technology, more and more sophisticated optical instruments have emerged, which also puts forward higher requirements for the field of image processing. In the field of fluorescence microscopy, different from the image quality problems caused by the random interference of hardware and light, the image blurring caused by the diffraction of light is the key research object of image processing. This kind of blurring can use a specific mathematical model to simulate the diffusion of light. The point spread function PSF can be theoretically defined through the mathematical model of diffraction and the parameters of the microscope.

[0003] In order to eliminate or reverse the blurring phenomenon of the image caused by light diffusion, a deconvolution algorithm is introduced. The deconvolution combines the PSF to invert the imaging convolution iteratively. Applying the deconvolution to image processing can improve the quality of the image, enhance the contrast and clarity of the image.

[0004] Existing traditional deconvolution algorithms have high restrictions on the image itself, are prone to loss of image information, and are not suitable for micro high-precision imaging. The deep learning algorithm based on neural network is computationally complex, has high requirements for device performance, and will cause waste of resources. Summary of the Invention

[0005] The purpose of the present application is to provide an image processing method, system and storage medium based on a deconvolution algorithm to solve one or more technical problems existing in the prior art, and at least provide a beneficial choice or create conditions.

[0006] The present application adopts the following technical solutions to achieve the above-mentioned invention purpose:

[0007] The present application provides an image processing method based on a deconvolution algorithm, including:

[0008] Obtaining the image data of a blurred image;

[0009] Generating a plurality of point spread functions that match through the optical microscopy system that captures the blurred image;

[0010] Selecting the best point spread function and a preset number of iterations from the plurality of point spread functions;

[0011] Preprocessing the selected best point spread function;

[0012] Performing Fourier transform on the point spread function through butterfly operation, and performing Fourier transform on the blurred image;

[0013] Perform deconvolution operation by means of iterative solution to obtain the target image;

[0014] Judge whether the iteration times are reached;

[0015] If so, output the final target image;

[0016] If not, use the target image obtained through iteration as the blurred image, and continue to perform repeated iteration until the iteration times are reached, and then output the final target image.

[0017] Furthermore, generate a number of corresponding point spread functions through the optical microscopy system that captures the blurred image, including:

[0018] In the optical microscopy system, collect different fluorescent microspheres and select N target fluorescent microspheres;

[0019] Construct N corresponding point spread functions h 1 (x, y), h 2 (x, y),......, h N (x, y) under this microscopy system, and apply them to the blurred image generated under this microscopy system.

[0020] Furthermore, select the best point spread function and the preset iteration times from a number of point spread functions, including:

[0021] Start from 1 and increase the iteration times i in turn. When the iteration times are i, use N point spread functions h 1 (x, y), h 2 (x, y),......, h N (x, y) to perform i iterations on the same blurred image;

[0022] Filter out the point spread functions where the target image is overexposed, over-noisy or loses details;

[0023] During the process of increasing the iteration times, if the target image is overexposed, over-noisy or loses details, stop increasing the iteration times;

[0024] Select the best point spread function h(x, y) and the iteration times k.

[0025] Furthermore, preprocess the selected best point spread function, including:

[0026] Fill the image around the point spread function with zero values to keep the image data of the point spread function centered, and generate a point spread function that matches the size of the blurred image.

[0027] Further, the Fourier transform of the point spread function and the Fourier transform of the blurred image are performed through butterfly operations, including:

[0028] The formula for one-dimensional discrete Fourier transform is as follows:

[0029]

[0030] Where The range of n is 0, 1, 2,..., N - 1, x(n) represents each sequence, Represents the fundamental frequency component of the sequence, Represents the k-th harmonic component of the sequence, the range of k is 0, 1, 2,..., N - 1, and X(k) represents the relative amplitude of the N discrete frequency points corresponding to the input sequence;

[0031] Through butterfly operations, the even part and the odd part in the original sequence of the point spread function are split, and the Fourier transform is respectively performed on the even part and the odd part, obtaining:

[0032]

[0033] Using the symbol 2r to represent even n and 2r + 1 to represent odd n, the range of r is 0, 1,..., N / 2 - 1, obtaining:

[0034]

[0035] Where Is the same coefficient for each odd component, let x 1 (r) = x(2r), x 2 (r) = x(2r + 1) to obtain the standard formula for butterfly operations:

[0036]

[0037] Where, x 1 (r) represents the odd sequence, x 2 (r) represents the even sequence, x 1 (r) and x 2 (r) both have only 2 / N sequences, Represents the rotation factor;

[0038] Extending this formula to two-dimensional Fourier transform, the fast Fourier transform formulas for the point spread function and the blurred image are:

[0039]

[0040] Further, an iterative solution method is used to perform deconvolution operation to obtain the target image, including:

[0041] Perform a complex multiplication on the Fourier transform result G(u, v) of the blurred image g(x, y) and the Fourier transform result H(u, v) of the point spread function h(x, y) to obtain a first intermediate variable. The formula is as follows:

[0042]

[0043] G(u, v) = a1 + b1*i

[0044] H(u, v) = a2 + b2*i

[0045] where U 1 (u, v) represents the first intermediate variable, (u, v) represents the coordinate variables in the frequency domain, G(u, v) represents the Fourier transform result of the blurred image, a1 represents the real part coefficient of the Fourier transform result of the blurred image, b1 represents the imaginary part coefficient of the Fourier transform result of the blurred image, i represents the imaginary unit, H(u, v) represents the Fourier transform result of the point spread function, a2 represents the real part coefficient of the Fourier transform result of the point spread function, and b2 represents the imaginary part coefficient of the Fourier transform result of the point spread function;

[0046] Perform an inverse Fourier transform on the first intermediate variable U 1 (u, v) to obtain u 1 (x, y), and divide the blurred image g(x, y) by the inverse transform result u 1 (x, y) to obtain a second intermediate variable. The formula is as follows:

[0047] u 1 (x, y) = IDFT(U 1 (u, v))

[0048]

[0049] Perform a Fourier transform on the second intermediate variable u 2 (x, y) to obtain U 2 (u, v), and perform a complex multiplication on the Fourier transform result U 2 (u, v) and the conjugate of the Fourier transform result of the point spread function to obtain a third intermediate variable U 3 (u, v). The formula is as follows:

[0050] U 2 (u, v) = DFT(u 2 (x, y))

[0051]

[0052] X(u, v) = a3 + b3*i

[0053]

[0054] Among them, U 3 (u, v) represents the third intermediate variable, U 2 (u, v) represents the second intermediate variable u 2 The Fourier transform result of (x, y), a3 represents the real part coefficient of the Fourier transform result of the second intermediate variable, and b3 represents the imaginary part coefficient of the Fourier transform result of the second intermediate variable. Represents the conjugate of the Fourier transform result H(u, v) of the point spread function, a4 represents the real part coefficient of the Fourier transform result of the second intermediate variable, and b4 represents the imaginary part coefficient of the Fourier transform result of the second intermediate variable;

[0055] Perform the inverse Fourier transform on the third intermediate variable U 3 (u, v) to obtain u 3 (x, y), and multiply the inverse Fourier transform result u 3 (x, y) with the blurred image g(x, y) to obtain the target image f(x, y). The formula is as follows:

[0056] u 3 (x, y) = IDFT(U 3 (u, v))

[0057] f(x, y) = u 3 (x, y)g(x, y).

[0058] This application provides an image processing system based on a deconvolution algorithm, including:

[0059] An acquisition unit for acquiring the image data of the blurred image;

[0060] A generation unit for generating a plurality of phase-matched point spread functions through an optical microscopy system that captures the blurred image;

[0061] A selection unit for selecting the best point spread function and a preset number of iterations from the plurality of point spread functions;

[0062] A preprocessing unit for preprocessing the selected best point spread function;

[0063] A first operation unit for performing a Fourier transform on the point spread function and a Fourier transform on the blurred image through a butterfly operation;

[0064] A second operation unit for performing a deconvolution operation by means of iterative solution to obtain the target image;

[0065] A judgment output unit for judging whether the number of iterations is reached;

[0066] If so, output the final target image;

[0067] If not, use the target image obtained through iteration as the blurred image, and continue to perform repeated iteration until the number of iterations is reached, and then output the final target image.

[0068] This application provides an image processing system based on a deconvolution algorithm, including a memory and a processor;

[0069] The memory is used to store instructions;

[0070] The processor is used to operate according to the instructions to execute the steps of the above method.

[0071] This application provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of the above method are implemented.

[0072] The beneficial effects of this application are as follows:

[0073] The deconvolution algorithm provided by this application combines the Fourier transform based on butterfly operations. Under the iterative idea, the blurred image is convolved, calculated, and then deconvolved to achieve the effect of image clarity. In order to simplify the calculation of convolution, this application uses the Fourier transform to convert frequency-domain convolution into time-domain multiplication, has low restrictions on the image itself, is not easy to lose image information, has low requirements for device performance, optimizes the algorithm, improves the operation efficiency, and reduces the burden on the device and resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 FIG. is an overall flowchart of an image processing method based on a deconvolution algorithm provided by an embodiment of this application;

[0075] Figure 2 FIG. is a specific flowchart of an image processing method based on a deconvolution algorithm provided by an embodiment of this application;

[0076] Figure 3 FIG. is a blurred image diagram in an image processing method based on a deconvolution algorithm provided by an embodiment of this application;

[0077] Figure 4 FIG. is a target image diagram in an image processing method based on a deconvolution algorithm provided by an embodiment of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0078] The present application discloses a deconvolution algorithm, which combines the Fourier transform based on butterfly operation and its inverse transform. Under the iterative idea, the blurred image is convolved, calculated and then deconvolved to achieve the effect of image clarity. In order to simplify the calculation of convolution, the present application uses the Fourier transform to convert the frequency-domain convolution into a time-domain product, which has low limitation on the image itself, is not easy to lose image information, has low requirements for device performance, optimizes the algorithm, improves the operation efficiency, and reduces the burden on devices and resources.

[0079] In existing traditional deconvolution algorithms, such as classical algorithms like inverse filtering, Wiener filtering, and least squares filtering, they all perform simple mathematical transformations on the frequency-domain image to obtain the target image. The algorithms are relatively simple but the deblurring effect is not good, and it is even more insufficient when used in the field of microscopic optics; while the deep learning algorithm based on neural network has a good deblurring effect, but it requires a large amount of data to train the model, has low operation efficiency, and the overall algorithm process is also relatively complex. The present application is based on iterative deconvolution, converting the time-domain convolution into a frequency-domain multiplication, which not only simplifies the operation process by mathematical calculation but also incorporates the iterative idea, effectively improving the deblurring effect.

[0080] Embodiment 1

[0081] As Figures 1 to 4 shown, a method for image processing based on a deconvolution algorithm provided by the present application specifically includes the following steps:

[0082] S100: Obtain the image data of the blurred image g(x, y);

[0083] S200: Generate a corresponding point spread function h(x, y) through the optical microscopy system that captures the blurred image g(x, y);

[0084] The specific content of S200 is as follows:

[0085] In the optical microscopy system, different fluorescent microspheres are collected, and N target fluorescent microspheres are selected. N corresponding point spread functions h 1 (x, y), h 2 (x, y),......, h N (x, y) are constructed through the image data of the target fluorescent microspheres, and are applied to the blurred image generated under this microscopy system;

[0086] S300: Conduct an experiment to select the best point spread function h(x, y) and the number of iterations k from the N point spread functions h 1 (x, y), h 2 (x, y),......, h N (x, y);

[0087] The specific content of S300 is as follows:

[0088] The selection of the point spread function h(x, y) and the number of iterations k are empirical values obtained from experiments. The specific experimental process is as follows: Starting from 1, the number of iterations i is increased sequentially. When the number of iterations is i, N point spread functions h 1 (x, y), h 2 (x, y),......, h N (x, y) are used to perform i iterations on the same blurred image, and the point spread functions for which the target image is overexposed, over-noisy, or loses details are filtered out; during the process of increasing the number of iterations, if the target image is overexposed, over-noisy, or loses details, the increase in the number of iterations is stopped. The best point spread function h(x, y) and the number of iterations k are selected by optimization.

[0089] S400: Preprocess the point spread function h(x, y) obtained in S300;

[0090] The specific content of S400 is as follows:

[0091] Process the image data of the point spread function h(x, y) so that the size of the point spread function h(x, y) matches that of the blurred image g(x, y). The specific processing method is to fill the image around h(x, y) with zero values and keep the image data of the point spread function centered;

[0092] S500: Perform Fourier transform on the point spread function h(x, y) to obtain H(u, v) through butterfly operation, and perform Fourier transform on the blurred image g(x, y) to obtain G(u, v);

[0093] The specific content of S500 is as follows:

[0094] First, the ordinary one-dimensional discrete Fourier transform DFT is as follows:

[0095]

[0096] Among them, The range of n is 0, 1, 2,..., N - 1, x(n) represents each sequence, represents the fundamental frequency component of the sequence, represents the k-th harmonic component of the sequence, the range of k is 0, 1, 2,..., N - 1, and X(k) represents the relative amplitude of the N discrete frequency points corresponding to the input sequence.

[0097] The butterfly operation is a fast algorithm for DFT. Through the butterfly operation, the even part and the odd part in the original sequence of the point spread function are split, and Fourier transform is performed on the even part and the odd part respectively:

[0098]

[0099] Represent the even number \(n\) by the symbol \(2r\) and the odd number \(n\) by \(2r + 1\). The range of \(r\) is \(0, 1,\cdots, N / 2 - 1\). There is:

[0100]

[0101] where is the same coefficient of each odd component. Let \(x\) 1 (r)=x(2r), \(x\) 2 (r)=x(2r + 1), and the standard formula for the butterfly operation can be obtained:

[0102]

[0103] where \(x\) 1 (r) represents the odd sequence, \(x\) 2 (r) represents the even sequence, \(x\) 1 (r) and \(x\) 2 (r) both have only \(2 / N\) sequences. represents the rotation factor.

[0104] Extending this formula to the two-dimensional Fourier transform, the point spread function and the fast Fourier transform formula of the blurred image can be obtained as:

[0105]

[0106] The point spread function and the Fourier transform of the blurred image can be simply expressed as:

[0107] H(u, v)=DFT(h(x, y))

[0108] G(u, v)=DFT(g(x, y))

[0109] Similarly, if you want to convert H(u, v) to h(x, y), it is the inverse Fourier transform, which is expressed as h(x, y)=IDFT(H(u, v)).

[0110] S600: Use the iterative solution method to perform deconvolution operation to obtain the target image f(x, y);

[0111] The overall idea of the iterative solution is specifically:

[0112] The first iterative process is as follows:

[0113]

[0114] Among them, * represents convolution, f(x, y) represents the target image, g(x, y) represents the blurred image, and h(x, y) represents the point spread function that matches the blurred image. According to the above formula, the target image f is obtained after the (k + 1)-th iteration. k+1 f k (x, y) is the target image at the k-th iteration. k f k (x, y) is multiplied by a coefficient related to f k (x, y) to obtain the target image f k (x, y) at the k-th iteration. k f k (x, y) is the target image at the k-th iteration. k f k (x, y) is the target image at the (k - 1)-th iteration. k-1 f k (x, y) is multiplied by a coefficient related to f k (x, y). k-1 And so on:

[0115]

[0116] Among them, * represents convolution, f(x, y) represents the target image, f k+1 (x, y) represents the target image at the (k + 1)-th iteration, g(x, y) represents the blurred image, h(x, y) represents the point spread function that matches the blurred image, and f k (x, y) represents the target image at the k-th iteration, and T represents the transpose operation of the matrix. k+1 f k+1 (x, y) represents the target image at the (k + 1)-th iteration, g(x, y) represents the blurred image, h(x, y) represents the point spread function that matches the blurred image, and f k (x, y) represents the target image at the k-th iteration, and T represents the transpose operation of the matrix. k f k (x, y) represents the target image at the k-th iteration, and T represents the transpose operation of the matrix.

[0117] For a single iteration, that is, generating f(x, y) from g(x, y) or generating f k (x, y) from f k (x, y), the time-domain convolution is equal to the frequency-domain multiplication. The image data in the time domain is converted into the values in the frequency domain through the Fourier transform. To facilitate the distinction of parameters, taking generating f(x, y) from g(x, y) as an example, the specific steps are as follows: k generating f k+1 (x, y) from f k+1 (x, y). k+1 For generating f(x, y) from g(x, y), the time-domain convolution is equal to the frequency-domain multiplication. The image data in the time domain is converted into the values in the frequency domain through the Fourier transform. To facilitate the distinction of parameters, taking generating f(x, y) from g(x, y) as an example, the specific steps are as follows:

[0118] First, perform the Fourier transform on the blurred image g(x, y) to obtain G(u, v).

[0119] G(u, v) = DFT(g(x, y))

[0120] Secondly, perform complex multiplication on the Fourier transform result G(u, v) of the blurred image g(x, y) and the Fourier transform result H(u, v) of the point spread function h(x, y) to obtain the first intermediate variable:

[0121]

[0122] G(u, v) = a1 + b1 * i

[0123] H(u, v) = a2 + b2 * i

[0124] Among them, U 1(u, v) represents the first intermediate variable, (u, v) represents the coordinate variable in the frequency domain, G(u, v) represents the Fourier transform result of the blurred image, a1 represents the real part coefficient of the Fourier transform result of the blurred image, b1 represents the imaginary part coefficient of the Fourier transform result of the blurred image, i represents the imaginary unit, H(u, v) represents the Fourier transform result of the point spread function, a2 represents the real part coefficient of the Fourier transform result of the point spread function, and b2 represents the imaginary part coefficient of the Fourier transform result of the point spread function;

[0125] Then, perform the inverse Fourier transform on the first intermediate variable U 1 (u, v) to obtain u 1 (x, y), and divide the blurred image g(x, y) by the inverse transform result u 1 (x, y) to obtain the second intermediate variable;

[0126] u 1 (x, y) = IDFT(U 1 (u, v))

[0127]

[0128] Next, perform the Fourier transform on the second intermediate variable u 2 (x, y) to obtain U 2 (u, v), and multiply the Fourier transform result U 2 (u, v) by the conjugate of the Fourier transform result of the point spread function to obtain the third intermediate variable U 3 (u, v):

[0129] U 2 (u, v) = DFT(u 2 (x, y))

[0130]

[0131] X(u, v) = a3 + b3 * i

[0132]

[0133] where U 3 (u, v) represents the third intermediate variable, U 2 (u, v) represents the Fourier transform result of the second intermediate variable u 2 (x, y), a3 represents the real part coefficient of the Fourier transform result of the second intermediate variable, and b3 represents the imaginary part coefficient of the Fourier transform result of the second intermediate variable, Denote the conjugate of the Fourier transform result H(u, v) of the point spread function, a4 represents the real part coefficient of the Fourier transform result of the second intermediate variable, and b4 represents the imaginary part coefficient of the Fourier transform result of the second intermediate variable;

[0134] Finally, perform the inverse Fourier transform on the third intermediate variable U 3 (u, v) to obtain u 3 (x, y), and multiply the inverse Fourier transform result u 3 (x, y) with the blurred image g(x, y) to obtain the target image f(x, y).

[0135] u 3 (x, y) = IDFT(U 3 (u, v))

[0136] f(x, y) = u 3 (x, y)g(x, y)

[0137] S700: Determine whether the preset number of iterations is reached; if so, output the final target image; otherwise, use the target image f k (x, y) obtained in this k-th iteration as the blurred image g(x, y), return to S600, and continue to perform repeated iterations until the preset number of iterations is reached, and output the target image f(x, y), that is, f k+1 (x, y) after k iterations.

[0138] Embodiment 2

[0139] This application also provides an image processing system based on the deconvolution algorithm, including:

[0140] An acquisition unit, configured to acquire the image data of the blurred image;

[0141] A generation unit, configured to generate a plurality of point spread functions that match through an optical microscopy system for capturing the blurred image;

[0142] A selection unit, configured to select the best point spread function and the preset number of iterations from the plurality of point spread functions;

[0143] A preprocessing unit, configured to preprocess the selected best point spread function;

[0144] A first operation unit, configured to perform the Fourier transform on the point spread function and the Fourier transform on the blurred image through butterfly operations;

[0145] A second operation unit, configured to perform deconvolution operations by using an iterative solution method to obtain the target image;

[0146] A judgment output unit, configured to judge whether the number of iterations is reached;

[0147] If so, output the final target image;

[0148] If not, use the target image obtained through iteration as the blurred image, and continue to perform repeated iteration until the number of iterations is reached, and then output the final target image.

[0149] Embodiment 3

[0150] Another image processing system based on the deconvolution algorithm provided by this application may further include a memory and a processor; the memory is used to store instructions;

[0151] The processor is used to operate according to the instructions to execute the steps of the foregoing image processing method based on the deconvolution algorithm.

[0152] Embodiment 4

[0153] This application also provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the steps of the foregoing image processing method based on the deconvolution algorithm.

[0154] Those skilled in the art should understand that the embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program codes.

[0155] This application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of this application. It should be understood that each process and / or block in the flowchart and / or block diagram, and the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate for implementing in the process Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks the device with the specified functions.

[0156] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured product including an instruction device, and the instruction device implements in the processFigure 1 one process or multiple processes and / or blocks Figure 1 the functions specified in one block or multiple blocks.

[0157] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0158] The above are only the preferred embodiments of the present application. It should be noted that for those of ordinary skill in the art of this technology, without departing from the technical principle of the present application, several improvements and modifications can also be made, and these improvements and modifications should also be regarded as the protection scope of the present application.

Claims

1. An image processing method based on a deconvolution algorithm, characterized in that: include: Obtain image data of the blurred image; A number of matched point spread functions are generated by an optical microscope system that captures the blurred image; Selecting an optimal point spread function and a preset number of iterations from a plurality of point spread functions; Preprocessing the selected optimal point spread function; Performing Fourier transform on the point spread function and the blurred image through butterfly operation; Using iterative solution method, deconvolution operation is performed to obtain the target image; Determine whether the number of iterations has been reached; If so, output the final target image; If not, the target image obtained after iteration is used as the blurred image, and the iteration is repeated until the number of iterations is reached, and the final target image is output.

2. The image processing method based on the deconvolution algorithm according to claim 1, characterized in that: Several point spread functions are matched by the optical microscope system that captures the blurred image, including: In an optical microscope system, different fluorescent microspheres are collected and N target fluorescent microspheres are selected; The corresponding N point spread functions h1(x,y), h2(x,y), ..., are constructed under the microscope system through the target fluorescent microsphere image data. h N (x,y) and applied to the blurred image generated under this microscope system.

3. The image processing method based on the deconvolution algorithm according to claim 1, characterized in that: Select the best point spread function and a preset number of iterations from several point spread functions, including: Starting from 1, the number of iterations i is increased successively. When the number of iterations is i, N point spread functions h1(x, y), h2(x, y), ..., are used respectively. h N (x,y) Iterate the same blurred image i times; Filter out the point spread function of the target image that is overexposed, noisy, or loses details; In the process of increasing the number of iterations, if the target image is overexposed, too noisy, or loses details, stop increasing the number of iterations; Select the best point spread function h(x,y) and number of iterations k.

4. The image processing method based on the deconvolution algorithm according to claim 1, characterized in that: Preprocessing of the selected optimal point spread function includes: The image around the point spread function is filled with zero values ​​to keep the point spread function image data centered and generate a point spread function that matches the size of the blurred image.

5. The image processing method based on the deconvolution algorithm according to claim 1, characterized in that: Perform Fourier transform on the point spread function and Fourier transform on the blurred image through butterfly operation, including: The formula for the one-dimensional discrete Fourier transform is as follows: in, The range of n is 0, 1, 2, ..., N-1, x(n) represents each sequence, represents the fundamental frequency component of the sequence, represents the kth harmonic component of the sequence, where k ranges from 0, 1, 2, ..., N-1, and X(k) represents the relative amplitude of the N discrete frequency points corresponding to the input sequence; By butterfly operation, the even part and the odd part in the original sequence of the point spread function are split, and Fourier transform is performed on the even part and the odd part respectively to obtain: Using the symbol 2r to represent an even number n, 2r+1 to represent an odd number n, and the range of r is 0, 1, ..., N / 2-1, we get: in, are the same coefficients of the odd components, and x1(r)=x(2r), x2(r)=x(2r+1) gives the standard formula for butterfly operation: Among them, x1(r) represents an odd sequence, x2(r) represents an even sequence, and there are only 2 / N sequences of x1(r) and x2(r). and represents the rotation factor; By extending this formula to the two-dimensional Fourier transform, we get the point spread function and the fast Fourier transform formula of the blurred image:

6. The image processing method based on the deconvolution algorithm according to claim 5, characterized in that: Using iterative solution, deconvolution operation is performed to obtain the target image, including: The Fourier transform result G(u,v) of the blurred image g(x,y) is complex multiplied with the Fourier transform result H(u,v) of the point spread function h(x,y) to obtain the first intermediate variable, as shown in the following formula: G(u,v)=a1+b1*i H(u,v)=a2+b2*i Wherein, U1(u,v) represents the first intermediate variable, (u,v) represents the coordinate variable in the frequency domain, G(u,v) represents the Fourier transform result of the blurred image, a1 represents the real coefficient of the Fourier transform result of the blurred image, b1 represents the imaginary coefficient of the Fourier transform result of the blurred image, i represents the imaginary unit, H(u,v) represents the Fourier transform result of the point spread function, a2 represents the real coefficient of the Fourier transform result of the point spread function, and b2 represents the imaginary coefficient of the Fourier transform result of the point spread function; Perform inverse Fourier transform on the first intermediate variable U1(u,v) to obtain u1(x,y), and divide the blurred image g(x,y) by the inverse transform result u1(x,y) to obtain the second intermediate variable. The formula is as follows: u1(x,y)=IDFT(U1(u,v)) Perform Fourier transform on the second intermediate variable u2(x,y) to obtain U2(u,v), and conjugate the Fourier transform result U2(u,v) with the Fourier transform result of the point spread function Perform complex multiplication to obtain the third intermediate variable U3(u,v), the formula is as follows: U2(u,v)=DFT(u2(x,y)) X(u,v)=a3+b3*i Wherein, U3(u,v) represents the third intermediate variable, U2(u,v) represents the Fourier transform result of the second intermediate variable u2(x,y), a3 represents the real coefficient of the Fourier transform result of the second intermediate variable, and b3 represents the imaginary coefficient of the Fourier transform result of the second intermediate variable. represents the conjugate of the Fourier transform result H(u,v) of the point spread function, a4 represents the real coefficient of the Fourier transform result of the second intermediate variable, and b4 represents the imaginary coefficient of the Fourier transform result of the second intermediate variable; Perform inverse Fourier transform on the third intermediate variable U3(u,v) to obtain u3(x,y), and multiply the inverse Fourier transform result u3(x,y) with the blurred image g(x,y) to obtain the target image f(x,y), as shown in the following formula: u3(x,y)=IDFT(U3(u,v)) f(x,y)=u3(x,y)g(x,y).

7. An image processing system based on a deconvolution algorithm, characterized in that: include: An acquisition unit, used for acquiring image data of a blurred image; A generating unit, for generating a plurality of matching point spread functions by an optical microscope system for capturing a blurred image; A selection unit, used for selecting an optimal point spread function and a preset number of iterations from a plurality of point spread functions; A preprocessing unit, used for preprocessing the selected optimal point spread function; A first operation unit, used for performing Fourier transform on the point spread function and Fourier transform on the blurred image through butterfly operation; The second operation unit is used to perform a deconvolution operation by an iterative solution method to obtain a target image; A judgment output unit is used to judge whether the number of iterations has been reached; If so, output the final target image; If not, the target image obtained after iteration is used as the blurred image, and the iteration is repeated until the number of iterations is reached, and the final target image is output.

8. An image processing system based on a deconvolution algorithm, characterized in that: including memory and processor; The memory is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images