Microscopic image distortion correction and reconstruction method, system, device and storage medium

Abstract

The present application provides a kind of microscopic image distortion correction and reconstruction method, system, equipment and storage medium, preset multi-shape calibration slide, including: regular grid slide, concentric circle slide, irregular curve slide and multi-layer depth slide;The image of multi-shape calibration slide under different field of view, different magnification and different focal plane is collected by microscope, and the geometric distortion and blur characteristics of image are quantified by combining edge detection, contour analysis and blur measurement method, and distortion feature vector is generated;Point spread function model is constructed based on distortion feature vector, and point spread function model is optimized using convolutional neural network;The optimized point spread function model is applied to the deconvolution processing and geometric correction of microscopic image, and finally the clear image after reconstruction is obtained.The present application can effectively correct geometric distortion and edge blur, restore the details and clarity of microscopic image.It has wide application prospect in scientific research, clinical diagnosis and industrial detection field.

Description

Technical Field

[0001] This application relates to the field of microscopic imaging technology, specifically to a method, system, device, and storage medium for microscopic image distortion correction and reconstruction. Background Technology

[0002] Microscopic imaging technology has important applications in biomedical research, materials science and industrial inspection, but due to limitations such as objective lenses, light sources, cameras and mechanical deviations, microscope optical systems often suffer from problems such as blurring and geometric distortion, which seriously affect image quality and the reliability of subsequent analysis.

[0003] Traditional PSF (Point Spread Function) modeling methods rely on calibration samples of a single geometry, such as standard grids or lattices. These methods struggle to fully capture complex distortion patterns, resulting in limited distortion correction capabilities. Specifically: Insufficient capture of distortion features: Existing PSF modeling methods rely on calibration samples of single points or simple meshes, which makes it difficult to capture complex distortion features, such as nonlinear geometric deformation and edge distortion.

[0004] The PSF model has weak generalization ability: the lack of diverse distortion training data makes it difficult for the traditional PSF model to adapt to different optical configurations and sample conditions, resulting in unstable correction results.

[0005] Limited image reconstruction accuracy: Existing correction techniques struggle to achieve consistent sharpness across the entire field of view, especially in high-resolution imaging where distortion in the edge and depth directions is particularly pronounced.

[0006] In view of this, this application proposes a method, system, device and storage medium for microscopic image distortion correction and reconstruction, which can significantly improve the clarity and accuracy of microscopic imaging. Summary of the Invention

[0007] To address the limitations of existing technologies in capturing complex distortion patterns and in correcting distortion in microscopic images, this application provides a method, system, device, and storage medium for microscopic image distortion correction and reconstruction, thereby resolving the aforementioned technical deficiencies.

[0008] In a first aspect, the present invention proposes a method for microscopic image distortion correction and reconstruction, the method comprising the following steps: S1. Preset multi-shape calibration slides, including: regular grid slides, concentric circle slides, irregular curve slides and multi-layer depth slides; S2. Images of multi-shaped calibration slides under different fields of view, magnification, and focal planes are acquired using a microscope. The geometric distortion and blur characteristics of the images are quantified by combining edge detection, contour analysis, and blur measurement methods to generate distortion feature vectors. S3. Construct a point spread function model based on the distorted feature vector, and optimize the point spread function model using a convolutional neural network; S4. Apply the optimized point spread function model to perform deconvolution processing and geometric correction on the microscopic image to finally obtain a clear reconstructed image.

[0009] Preferably, in step S1, the line width of the grid on the regular grid slide is 5µm±0.5µm, and the spacing between two grid lines is 50µm±5µm; The concentric glass slide contains concentric circles with radii ranging from 5µm to 100µm and distributed dots; Irregular curve slides consist of random curves and a distributed dot matrix; The multi-layer depth slide contains multiple transparent planes, each marked with a different pattern, and the distance between two adjacent transparent planes is 5µm±0.2µm.

[0010] Preferably, in step S2, the geometric distortion and blur characteristics of the image are quantified by combining edge detection, contour analysis, and blur measurement methods to generate a distortion feature vector, specifically including the following sub-steps: S21. After denoising the acquired image, perform edge detection on the image, extract and analyze the contours of the edge detection results, and obtain key features. S22. Obtain the actual geometric values ​​of the multi-shape calibration slide, and calculate the geometric distortion measure of the image based on the actual geometric values ​​and key features; S23. Calculate the intensity change of the image using the image gradient and Laplacian transform methods to obtain the image blur. S24. Integrate geometric distortion measure and ambiguity into a distortion feature vector.

[0011] More preferably, in step S22, the geometric distortion measure of the image includes: stretching amount, rotation angle, distortion degree, edge sharpness, and dot spread range; The expression for calculating geometric distortion metric is: ×100%.

[0012] More preferably, in step S23, the expression for calculating the image gradient magnitude is:

[0013] In the formula, This represents the gradient of the image in the horizontal direction (x-direction); This represents the gradient of the image in the vertical direction (y-direction); This represents the magnitude of the image gradient, used to quantify changes in image intensity. The expression for calculating the Laplace variance is:

[0014] In the formula, This represents the Laplacian transform result of an image, used for detecting edges and details in the image; Variance represents the variance, used to quantify the degree of dispersion of the Laplacian transform result, reflecting the blurring characteristics of the image.

[0015] Preferably, in step S3, a convolutional neural network is used to optimize the point spread function model, wherein the convolutional neural network includes: The input layer is used to receive image data from multi-shape calibration slides. Five convolutional layers, each using a 3×3 convolutional kernel to extract local distortion features; Two pooling layers are used, with a 2×2 pooling window and a stride of 2 for feature map dimensionality reduction. The system consists of three fully connected layers, which map the distortion features to the point spread function model parameters.

[0016] Preferably, in step S4, the optimized point spread function model is applied to perform deconvolution processing and geometric correction on the microscopic image to finally obtain a reconstructed clear image, including the following steps: S41. Obtain microscopic images; S42. Use the optimized point spread function model as the core parameter in the deconvolution process; S43. Using Wiener deconvolution or Lucy-Richardson deconvolution algorithms, the microscopic image is the result of convolution between a clear image and a point spread function model. The undistorted image is recovered through an iterative process. Furthermore, based on the geometric data of the multi-shape calibration slide, affine transformation is performed to correct the proportions and shape, ultimately obtaining a clear image.

[0017] Secondly, this invention proposes a microscopic image distortion correction and reconstruction system, which includes: A multi-shape calibration slide design module is configured with preset multi-shape calibration slides, which include: regular grid slides, concentric circle slides, irregular curve slides, and multi-layer depth slides. The feature vector generation module is configured to generate distortion feature vectors by combining edge detection, contour analysis and blur measurement methods to quantify the geometric distortion and blur characteristics of the images through images of multi-shaped calibration slides acquired by microscopes under different fields of view, magnification and focal plane. The model optimization module is configured to construct a point spread function model based on distorted feature vectors and optimize the point spread function model using a convolutional neural network. The output module is configured to apply an optimized point spread function model to perform deconvolution processing and geometric correction on the microscopic image, ultimately obtaining a clear reconstructed image.

[0018] Thirdly, the present invention proposes a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the microscopic image distortion correction and reconstruction method as described in any of the preceding claims.

[0019] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the microscopic image distortion correction and reconstruction method as described in any of the preceding claims.

[0020] Compared with the prior art, the beneficial results of the present invention are as follows: (1) Comprehensive Distortion Correction: The PSF model trained with multi-shaped slides effectively corrects geometric distortion and edge blurring. This invention captures the multidimensional distortion features of the microscope system, including nonlinear geometric deformation and edge distortion, using regular grid slides, concentric circle slides, irregular curve slides, and multi-layer depth slides. This significantly improves the ability to capture distortion features and provides rich data support for subsequent PSF modeling.

[0021] (2) High-precision image reconstruction: The deconvolution algorithm combined with the optimized PSF model restores the details and clarity of the microscopic image. This invention optimizes the PSF model using deep learning technology, especially convolutional neural networks (CNN), significantly enhancing the model's generalization ability, enabling it to adapt to various optical configurations and sample conditions, and improving the stability of the correction effect. Especially in high-resolution imaging, it can effectively correct distortion problems in the edge and depth directions, achieving consistent clarity improvement across the entire field of view.

[0022] (3) Multi-scene adaptability: The PSF model is adaptable to different microscope configurations and optical systems, and has wide applicability. Through the optimized PSF model and deconvolution algorithm, this invention can quickly and accurately correct and reconstruct microscopic images, which significantly improves the efficiency and reliability of microscopic imaging and provides more accurate image data support for scientific research and industrial inspection. Attached Figure Description

[0023] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments, taken with reference to the accompanying drawings: Figure 1 This is a flowchart of the microscopic image distortion correction and reconstruction method according to this application; Figure 2a This is a schematic diagram of a glass slide with a regular grid pattern according to this application; Figure 2b This is a schematic diagram of a concentric glass slide according to this application; Figure 2c This is a schematic diagram of an irregularly curved glass slide according to this application; Figure 3 This is a structural diagram of the microscopic image distortion correction and reconstruction system according to this application; Figure 4 This is a schematic diagram of the structure of a computer system suitable for implementing the electronic devices of the present application embodiments. Detailed Implementation

[0024] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0025] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0026] Figure 1 A flowchart of the microscopic image distortion correction and reconstruction method of this application is shown. Please refer to it. Figure 1 The method includes the following steps: This invention proposes a method for microscopic image distortion correction and reconstruction, which includes the following steps: S1. Preset multi-shape calibration slides, including: regular grid slides, concentric circle slides, irregular curve slides and multi-layer depth slides.

[0027] Figure 2a A schematic diagram of the regular grid slide of this application is shown, as follows: Figure 2a As shown, the line width of the grid on the regular grid slide is 5µm ± 0.5µm, and the spacing between two grid lines is 50µm ± 5µm. The preferred line width is 5µm, and the spacing is 50µm. This regular grid slide is used for overall geometric distortion (e.g., scaling, stretching, twisting).

[0028] In this embodiment, a regular grid slide design with a line width of 5µm and a spacing of 50µm was chosen primarily to match the optical resolution limit of the microscope. According to the Nyquist sampling theorem, resolution is typically calculated based on:

[0029] in, λ is the wavelength of light (commonly 550 nm), and NA is the numerical aperture of the objective lens.

[0030] For different objectives, the resolution is calculated as follows: 20X0.5NA: ≈0.671µm 20X0.8NA: ≈0.42375µm 40X0.75NA: ≈0.449µm 20X0.95NA: ≈0.346µm Choosing a line width of 5µm is far greater than the resolution limit of different objectives, while ensuring image clarity. A spacing of 50µm ensures sufficient contrast between lines when observed under a microscope, meeting practical application requirements.

[0031] Figure 2b A schematic diagram of the concentric glass slide of this application is shown, as follows: Figure 2b As shown, the concentric circular slide comprises concentric circles with radii ranging from 5µm to 100µm and distributed dots. This concentric circular slide is used to detect radial and tangential distortion, as well as edge distortion.

[0032] Figure 2c A schematic diagram of the irregularly curved glass slide of this application is shown, as follows: Figure 2c As shown, the irregular curve slide consists of random curves and a distributed lattice to simulate complex distortions. This irregular curve slide is used to enhance the adaptability of the PSF model to non-standard distortions.

[0033] The multi-layer depth slide comprises multiple layers of transparent planes, each marked with a different pattern, and the distance between adjacent transparent planes is 5µm ± 0.2µm. A distance of 5µm is preferred. This multi-layer depth slide is used to capture focal plane shift (depth distortion) in the Z-axis direction.

[0034] In this embodiment, the method for determining the 5µm spacing between transparent planes (based on objective depth of focus or Z-axis scanning accuracy) is as follows: The setting of 5µm for the transparent plane spacing is based on the depth of focus (DOF) of the objective lens and the Z-axis scanning accuracy. The expression for calculating the depth of focus (DOF) is as follows: DOF=

[0035] In the formula, λ is the wavelength of light, NA is the numerical aperture of the objective lens, and n is the refractive index of the medium (e.g., the refractive index of the medium on the slide or sample).

[0036] The Z-axis scanning repeatability is ≤±0.2µm. In terms of accuracy requirements, the design spacing of 5µm helps to ensure that the image layers are distinct and do not overlap when performing Z-axis scanning.

[0037] Combining the lens parameters above, the following depth-of-focus results can be obtained: 20 x 0.5 NA: DOF ≈ 1.25 µm 20 x 0.8 NA: DOF ≈ 0.55 µm 40 x 0.75 NA: DOF ≈ 0.33 µm 20 x 0.95 NA: DOF ≈ 0.23 µm Therefore, this application selects a transparent plane spacing of 5µm, which can effectively cover the microscope's depth of focus and meet the requirements for Z-axis repeatability.

[0038] Continue to refer to Figure 1 The microscopic image distortion correction and reconstruction method proposed in this application further includes the following steps: S2. Images of multi-shaped calibration slides under different fields of view, magnification, and focal planes are acquired using a microscope. The geometric distortion and blur characteristics of the images are quantified by combining edge detection, contour analysis, and blur measurement methods to generate distortion feature vectors.

[0039] In this embodiment, the following sub-steps are specifically included: S21. After denoising the acquired image, perform edge detection on the image, extract and analyze the contours of the edge detection results, and obtain key features.

[0040] Preferably, background noise is removed using methods such as Gaussian filtering to enhance image quality. The Canny or Sobel algorithm is then used to perform edge detection on the denoised image to identify key features. Image processing tools (OpenCV) are then used to extract and analyze contour information to obtain key features, including geometric features such as shape, area, and perimeter.

[0041] S22. Obtain the actual geometric values ​​of the multi-shape calibration slide, and calculate the geometric distortion measure of the image based on the actual geometric values ​​and key features. The geometric distortion measure of the image includes: stretching, rotation angle, distortion, edge sharpness, and point spread. The expression for calculating geometric distortion metric is: ×100%.

[0042] Specifically, stretching: calculates the proportional difference between a standard shape and the actual shape in an image. It can be quantified by comparing the measured actual geometry of a gridded slide with the expected geometry (e.g., a regular grid). The formula is:

[0043] To obtain the actual geometric value, These are the expected geometric eigenvalues; Rotation angle: The difference between the correct defined direction and the actual direction of the image center point can be calculated using the Hough transform or a similar method.

[0044] Distortion: Quantified by calculating the deviation of each point's position (usually the root mean square of the deviation) by comparing it with a standard grid.

[0045] Edge sharpness: quantified by calculating the edge intensity and contrast in the image, using the Laplacian operator to determine the edge response function.

[0046] In the formula This represents the i-th pixel or region in the image; This represents the Laplacian operator, used to detect edges in an image; N represents the total number of pixels or regions. This indicates edge sharpness; a higher value indicates sharper edges.

[0047] Point Spread Function (PSF): Calculates the severity of point spread within a specific focal plane and measures the spread width at key points.

[0048] The above indicators are integrated into a feature vector, in the form of:

[0049] Each subvector represents a different calibration slide or a different field of view.

[0050] Geometric distortion metric: Deviation: Calculates the distance or deviation between the actual image and the reference image at each point, typically using Euclidean distance.

[0051] In the formula , This represents the coordinates in the actual image; , Represents the coordinates in the theoretical (reference) image; It represents the Euclidean distance between actual and theoretical coordinates, and is used to quantify geometric distortion.

[0052] Furthermore, the deviation rate is calculated using the following expression: ×100% S23. The intensity change of the image is calculated using the image gradient and Laplacian transform methods to obtain the blur of the image.

[0053] The degree of blur can be quantified by calculating the image gradient and using methods such as the Laplacian transform.

[0054] Image intensity changes are calculated using methods such as image gradient and Laplacian transform. The blurriness is then calculated using the following formula:

[0055] in Is the image at a point? Strength at that location, This represents the gradient operation of the image (using Gaussian filtering).

[0056] In a specific embodiment, in step S23, the expression for calculating the image gradient magnitude is:

[0057] In the formula, This represents the gradient of the image in the horizontal direction (x-direction); This represents the gradient of the image in the vertical direction (y-direction); This represents the magnitude of the image gradient, used to quantify changes in image intensity. The expression for calculating the Laplace variance is:

[0058] In the formula, This represents the Laplacian transform result of an image, used for detecting edges and details in the image; Variance represents the variance, used to quantify the degree of dispersion of the Laplacian transform result, reflecting the blurring characteristics of the image.

[0059] S24. Integrate geometric distortion measure and ambiguity into a distortion feature vector.

[0060] By integrating quantified data into feature vectors, a distortion distribution model of the microscope optical system is generated using regression models or machine learning methods (such as support vector machines), providing basic data for PSF model training.

[0061] Continue to refer to Figure 1 The microscopic image distortion correction and reconstruction method provided in this application also includes the following steps: S3. Construct a point spread function model based on the distorted feature vector, and optimize the point spread function model using a convolutional neural network.

[0062] In this embodiment, the obtained distortion feature vector is input into the point spread function modeling module to calculate the PSF function. Based on known optical principles and quantization data, a preliminary model of the PSF is constructed. The preliminary PSF function formula is:

[0063] in The point spread function (PSF) of an imaging system describes the degree of blurring produced by a light source in the imaging system. This is the error term.

[0064] PSF expression:

[0065] In the formula, e is the base of the natural logarithm (approximately 2.71828); x and y are the coordinates in the graph plane; is the standard deviation of the Gaussian distribution, used to control the diffusion degree of PSF.

[0066] In this embodiment, the PSF model parameters are dynamically adjusted through multiple rounds of training to more accurately adapt to complex distortion patterns. This includes the following steps: 1. Establish a Convolutional Neural Network (CNN) architecture: Design a CNN model based on convolutional, pooling, and fully connected layers to process the input distorted data. Define the input layer as the distorted data of multi-shape calibration slides, and the output layer as the optimized PSF parameters.

[0067] Specifically, it includes the following: The input layer is used to receive image data from multi-shape calibration slides. Five convolutional layers, each using a 3×3 convolutional kernel to extract local distortion features; Two pooling layers are used, with a 2×2 pooling window and a stride of 2 for feature map dimensionality reduction. The system consists of three fully connected layers, which map the distortion features to the point spread function model parameters.

[0068] 2. Training data preparation: Divide the dataset into training and validation sets to ensure the model can generalize. Use labeled data (ideal PSF and distorted images) to guide model training.

[0069] 3. Model Training: Forward propagation: Input the training data into the CNN, calculate the output, compare it with the actual labels, and calculate the loss (such as mean squared error).

[0070] Backpropagation: Using gradient descent, adjust the model parameters (weights and biases) to minimize the loss function. The Adam or SGD optimizer can be used for training.

[0071] Multiple iterations: Repeat the forward and backward propagation process until the model converges or reaches the predetermined number of iterations.

[0072] 4. Parameter adjustment and optimization: During training, monitor the loss on both the training and validation sets to prevent overfitting. Regularization techniques (such as Dropout) or early stopping can be used to optimize model performance.

[0073] Continue to refer to Figure 1 The microscopic image distortion correction and reconstruction method proposed in this application further includes the following steps: S4. Apply the optimized point spread function model to perform deconvolution processing and geometric correction on the microscopic image to finally obtain a clear reconstructed image.

[0074] In this embodiment, a blurred microscopic image to be processed can be obtained from a microscope experiment and used as input to the deconvolution algorithm. An optimized PSF model is applied: the trained optimized PSF is used as the core parameter in the deconvolution process.

[0075] Deconvolution algorithms (such as Wiener deconvolution or Lucy-Richardson deconvolution) are used to reconstruct sharp images from blurry images. The specific steps involve treating the blurry microscopic image as a convolution of the sharp image and a PSF (Power-Side Array), and then recovering the undistorted sharp image through a reasonable iterative process. Wiener deconvolution is suitable for low-noise environments (signal-to-noise ratio > 30dB); Lucy-Richardson deconvolution is suitable for high-noise environments, requiring the setting of the number of iterations (e.g., 50) and a convergence threshold (e.g., 1e-5).

[0076] Convolution operation: Suppose the blurred image I is the convolution of the sharp image J and PSF:

[0077] Where * represents the convolution operation. It's noise.

[0078] Deconvolution reconstruction: using analytical or iterative methods, a clear image J is recovered from a blurred microscopic image, so that the convolution result matches the original blurred image as closely as possible.

[0079] Using the geometric data of the calibration slider, an affine transformation is performed on the extracted image to restore the true proportions and shape, ensuring that the geometric properties of the image are consistent with the actual sample.

[0080] In addition, by using focus data from multiple depth slides, a 3D image can be reconstructed and the focal plane drift problem can be corrected to obtain a more accurate 3D structural representation.

[0081] Further reference Figure 3 As a implementation of the above method, in a second aspect, this application provides an embodiment of a microscopic image distortion correction and reconstruction system 300, which can be specifically applied to various electronic devices. The system 300 includes the following modules: The multi-shape calibration slide design module 310 is configured on a preset multi-shape calibration slide, which includes: a regular grid slide, a concentric circle slide, an irregular curve slide, and a multi-layer depth slide. The feature vector generation module 320 is configured to generate distortion feature vectors by combining edge detection, contour analysis and blur measurement methods to quantify the geometric distortion and blur characteristics of the images and acquire images of multi-shaped calibration slides under different fields of view, magnification and focal planes through a microscope. The model optimization module 330 is configured to construct a point spread function model based on distorted feature vectors and optimize the point spread function model using a convolutional neural network; The output module 340 is configured to apply the optimized point spread function model to perform deconvolution processing and geometric correction on the microscopic image, and finally obtain a clear reconstructed image.

[0082] Thirdly, the present invention proposes a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the microscopic image distortion correction and reconstruction method as described in any of the preceding claims.

[0083] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the microscopic image distortion correction and reconstruction method as described in any of the preceding claims.

[0084] The following is for reference. Figure 4 It shows a schematic diagram of the structure of a computer system 400 suitable for implementing terminal devices or servers in the embodiments of this application. Figure 4 The terminal device or server shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0085] like Figure 4 As shown, the computer system 400 includes a central processing unit (CPU) 401, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 402 or programs loaded from storage section 408 into random access memory (RAM) 403. The RAM 403 also stores various programs and data required for the operation of the system 400. The CPU 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0086] The following components are connected to I / O interface 405: input section 406 including keyboard, mouse, etc.; output section 407 including liquid crystal display (LCD) and speakers, etc.; storage section 408 including hard disk, etc.; and communication section 409 including network interface card such as LAN card, modem, etc. Communication section 409 performs communication processing via a network such as the Internet. Drive 410 is also connected to I / O interface 405 as needed. Removable media 411, such as disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 410 as needed so that computer programs read from them can be installed into storage section 408 as needed.

[0087] Specifically, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 409, and / or installed from removable medium 411. When the computer program is executed by central processing unit (CPU) 401, it performs the functions defined in the methods of this application. It should be noted that the computer-readable medium described in this application can be a computer-readable signal medium or a computer-readable medium or any combination thereof. The computer-readable medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, a computer-readable medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable medium that can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. Program code contained on a computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0088] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0089] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0090] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A method for microscopic image distortion correction and reconstruction, characterized in that, Includes the following steps: S1. Preset multi-shape calibration slides, the multi-shape calibration slides including: regular grid slides, concentric circle slides, irregular curve slides and multi-layer depth slides; S2. Images of the multi-shaped calibration slides under different fields of view, magnifications, and focal planes are acquired using a microscope. The geometric distortion and blur characteristics of the images are quantified using edge detection, contour analysis, and blur measurement methods to generate distortion feature vectors. This process includes the following sub-steps: S21. After denoising the acquired image, edge detection is performed on the image, and contour extraction and analysis are performed on the edge detection results to obtain key features; S22. Obtain the actual geometric values ​​of the multi-shape calibration slide, and calculate the geometric distortion metric of the image based on the actual geometric values ​​and the key features; S23. The intensity change of the image is calculated using the image gradient and Laplacian transform method to obtain the blur of the image; S24. Integrate the geometric distortion metric and the ambiguity into the distortion feature vector; S3. Construct a point spread function model based on the distortion feature vector, and optimize the point spread function model using a convolutional neural network; wherein the convolutional neural network includes: An input layer is used to receive image data from the multi-shape calibration slide; Five convolutional layers, each using a 3×3 convolutional kernel to extract local distortion features; Two pooling layers, wherein the pooling layers use a 2×2 pooling window and a stride of 2 to reduce the dimensionality of the feature map; Three fully connected layers, which map the distortion features to point spread function model parameters; S4. Apply the optimized point spread function model to perform deconvolution processing and geometric correction on the microscopic image to finally obtain a clear reconstructed image.

2. The method for microscopic image distortion correction and reconstruction according to claim 1, characterized in that, In step S1, the line width of the grid on the regular grid slide is 5µm±0.5µm, and the spacing between two grid lines is 50µm±5µm. The concentric glass slide comprises concentric circles with radii ranging from 5µm to 100µm and distributed dots; The irregular curve slide is composed of random curves and a distributed dot matrix; The multi-layer depth slide comprises multiple transparent planes, each marked with a different pattern, and the distance between two adjacent transparent planes is 5µm±0.2µm.

3. The method for microscopic image distortion correction and reconstruction according to claim 1, characterized in that, In step S22, the geometric distortion measures of the image include: stretching amount, rotation angle, distortion degree, edge sharpness, and dot spread range; The expression for calculating the geometric distortion metric is as follows: ×100%。 4. The method for microscopic image distortion correction and reconstruction according to claim 1, characterized in that, In step S23, the expression for calculating the image gradient magnitude is: In the formula, This represents the gradient of the image in the horizontal direction; This represents the gradient of the image in the vertical direction; This represents the magnitude of the image gradient, used to quantify changes in image intensity. The expression for calculating the Laplace variance is: In the formula, This represents the Laplacian transform result of an image, used for detecting edges and details in the image; Variance represents the variance, used to quantify the degree of dispersion of the Laplacian transform result, reflecting the blurring characteristics of the image.

5. The method for microscopic image distortion correction and reconstruction according to claim 1, characterized in that, In step S4, the optimized point spread function model is applied to the microscopic image for deconvolution processing and geometric correction, ultimately obtaining a reconstructed, clear image, including the following steps: S41. Obtain microscopic images; S42. Use the optimized point spread function model as the core parameter in the deconvolution process; S43. Using Wiener deconvolution or Lucy-Richardson deconvolution algorithms, the microscopic image is regarded as the convolution result of a clear image and a point spread function model. The undistorted image is recovered through an iterative process. Furthermore, based on the geometric data of the multi-shape calibration slide, an affine transformation is performed to correct the proportions and shape, and finally a clear image is obtained.

6. A microscopic image distortion correction and reconstruction system, characterized in that, The system includes: A multi-shape calibration slide design module is configured on a preset multi-shape calibration slide, which includes: a regular grid slide, a concentric circle slide, an irregular curve slide, and a multi-layer depth slide; The feature vector generation module is configured to generate distortion feature vectors by quantifying the geometric distortion and blur characteristics of the images of the multi-shaped calibration slides acquired through a microscope at different fields of view, magnifications, and focal planes, and by combining edge detection, contour analysis, and blur measurement methods. Specifically, it includes the following sub-steps: S21. After denoising the acquired image, edge detection is performed on the image, and contour extraction and analysis are performed on the edge detection results to obtain key features; S22. Obtain the actual geometric values ​​of the multi-shape calibration slide, and calculate the geometric distortion metric of the image based on the actual geometric values ​​and the key features; S23. The intensity change of the image is calculated using the image gradient and Laplacian transform method to obtain the blur of the image; S24. Integrate the geometric distortion measure and the ambiguity into the distortion feature vector; A model optimization module is configured to construct a point spread function model based on the distorted feature vector and optimize the point spread function model using a convolutional neural network; wherein the convolutional neural network includes: An input layer is used to receive image data from the multi-shape calibration slide; Five convolutional layers, each using a 3×3 convolutional kernel to extract local distortion features; Two pooling layers, wherein the pooling layers use a 2×2 pooling window and a stride of 2 to reduce the dimensionality of the feature map; The system consists of three fully connected layers that map the distortion features to point spread function model parameters; and an output module configured to apply the optimized point spread function model to perform deconvolution processing and geometric correction on the microscopic image, ultimately obtaining a reconstructed, clear image.

7. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the microscopic image distortion correction and reconstruction method as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the microscopic image distortion correction and reconstruction method as described in any one of claims 1 to 5.

Images