Fourier laminated microscopic imaging vignetting correction method based on adaptive Gaussian filtering
Through the adaptive Gaussian filtering method, the image standard deviation and average value are analyzed, the optimal Gaussian parameters are selected, the correction lookup table is generated, and the spectrum is updated iteratively, which solves the image inhomogeneity caused by the vignetting effect and improves the resolution and accuracy of Fourier stacked imaging.
Patent Information
- Application Number
- CN202510586131.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-08
- Publication Date
- 2025-08-08
AI Technical Summary
The existing Fourier stacked imaging technology is affected by vignetting in large field of view imaging, resulting in reduced brightness at the edge of the image, resulting in uneven brightness, affecting imaging resolution and accuracy.
Adaptive Gaussian filtering method is used to set the threshold by analyzing the standard deviation and mean value of the image, select the optimal Gaussian parameters, perform Fourier transform and inverse transform, generate a correction lookup table, iteratively update the image spectrum, remove the vignetting effect, and reconstruct a high-resolution uniform image.
Effectively removes the influence of vignetting effect, improves the quality and resolution of image reconstruction, and improves the uneven image intensity phenomenon.
Smart Images
Figure CN120447187A_ABST
Abstract
Description
(1) Technical field
[0001] The present invention belongs to the technical field of digital microscopic imaging, and in particular relates to a Fourier stack microscopic imaging vignetting correction method based on adaptive Gaussian filtering. (2) Background technology
[0002] Fourier stacking imaging technology is a new type of computational imaging technology. The core of the imaging technology lies in its combination of microscopic optical computational imaging and image processing. By combining traditional field-of-view imaging with quantitative phase recovery of Fourier transform, it overcomes the resolution limitations of traditional optical microscopes. It can not only reconstruct and restore the high-resolution intensity information of the sample, but also restore the phase information, ultimately achieving large-field-of-view and high-resolution images.
[0003] However, vignetting is an unavoidable problem in practical imaging systems, especially in wide-field imaging. Vignetting can reduce image edge brightness, creating uneven brightness and introducing wrinkle artifacts during reconstruction, which in turn affects image resolution. This directly reduces imaging accuracy and reliability.
[0004] Reducing the impact of the vignetting effect on the image can often improve the quality of the subsequent stacked and reconstructed image. Therefore, a Gaussian filter is adaptively selected for the read image to extract the vignetting from the target image, and the image information is retained to the maximum extent by referring to the standard deviation and average value of the currently corrected image to evaluate the intensity of the selected filter to obtain the best vignetting background, so that the reconstructed image can remove the severe intensity unevenness caused by the vignetting effect and achieve high-quality imaging under the influence of the vignetting effect. (3) Summary of the invention
[0005] The purpose of the present invention is to provide a Fourier stack microscopy imaging vignetting correction method based on adaptive Gaussian filtering, aiming to solve the problem of severe uneven image intensity of microscopic stack reconstruction caused by the influence of vignetting effect.
[0006] To achieve the above object, the present invention provides a method for correcting vignetting in Fourier stack microscopy imaging based on adaptive Gaussian filtering, the method comprising the following steps:
[0007] S1: Set up a digital microscope optical path to collect the sample to be tested, record the multi-angle intensity map of the sample to be tested, and perform correction and reconstruction based on the intensity map;
[0008] S2: Set a threshold value that highly retains image information based on the standard deviation and mean of the current image, and obtain the optimal Gaussian parameter by testing σ that meets the threshold condition;
[0009] S3: The images are sequentially converted to the frequency domain by FFT and Gaussian filtering is used to obtain the vignetting frequency domain information, and the information is normalized and accumulated to obtain the correction lookup table;
[0010] S4: Subtract the IFT correction lookup table from the original image to obtain the corrected image;
[0011] S5: After correction, take the intermediate illumination as the target image and add the initial phase to the frequency domain. The phase remains unchanged. Use the image under the corresponding illumination angle to update the amplitude part to obtain the updated image light field.
[0012] S6: Perform Fourier transform on the updated image light field to obtain an updated spectrum, and add the updated spectrum components within the corresponding sub-aperture to the initial image spectrum;
[0013] S7: Repeat steps S5 to S6 until convergence, that is, minimizing the amplitude of the model-predicted reconstructed image and the true measured image.
[0014] In S1, the digital microscope optical path is a Fourier stacked microscope imaging system optical path based on flat-panel LED array illumination.
[0015] In S2, by analyzing the standard deviation and mean of the image, the intensity range of the vignetting effect can be determined. The threshold setting filters out the areas most affected by vignetting and retains useful image information, thereby achieving adaptive adjustment and selecting the optimal Gaussian parameters, which can better match the spatial distribution characteristics of vignetting. The Gaussian filter modeling is expressed as:
[0016] G i (x,y)=exp(-((x i -u) 2 +(y i -v) 2 ) / 2*σ 2 )
[0017] Where (u,v) represents the center of the image, (x i ,y i ) is the pixel coordinate on the i-th image, σ is the standard deviation, which controls the width of the Gaussian function. The optimal matching parameters within the range are obtained through testing and comparison.
[0018] In S3, MATLAB reads the collected image, performs FFT to obtain frequency domain information, applies Gaussian filtering to obtain vignetting information, normalizes it, and accumulates it to generate a correction lookup table (LUT). It can be expressed as:
[0019]
[0020] Among them I i Indicates the i-th image read, F(I i) indicates a Fourier transform. The above indicates that after the image is Gaussian filtered, an inverse Fourier transform is performed to obtain the average to obtain the final LUT.
[0021] In S4, the LUT obtained above is applied to perform correction processing on the multi-angle vignetting images in the spatial domain.
[0022] In S5, the corrected image is reconstructed. During the reconstruction process, the image is transferred to the frequency domain and the phase is estimated by updating the amplitude, which is expressed as:
[0023]
[0024] represents the updated target complex amplitude distribution corresponding to the LED in the mth row and nth column, is the low-resolution image captured corresponding to the LED unit in the mth row and nth column. The superscript u represents the updated target spectrum and target complex amplitude, and c represents the captured image. It is called the initial target complex amplitude.
[0025] In S6, the spectrum of the updated target complex amplitude image is obtained by Fourier transform, and this low-resolution spectrum is used to update the spectrum components within the corresponding sub-aperture in the high-resolution spectrum of the object. The update formula is as follows:
[0026]
[0027] in The updated aperture spectrum is obtained by subtracting the initial spectrum from the updated target complex amplitude distribution spectrum corresponding to the LED in the mth row and nth column.
[0028] In S7, the reconstructed image is repeatedly iterated. The following is the error of the image reconstruction after a single iteration. The error is accumulated in each iteration, and then the designed error threshold is compared to determine whether the iteration is converged. The convergence formula is:
[0029]
[0030] Among them, Loss n (O,P) represents the loss function of the nth iteration, where O is the complex amplitude of the sample, I n represents the low-resolution image captured at the n-th illumination angle in the n-th iteration.
[0031] The present invention designs a vignetting correction method for Fourier stack microscopy imaging based on adaptive Gaussian filtering. The present invention sets a threshold value for highly retaining the original image according to the standard deviation and average value of the current image, and obtains the optimal Gaussian parameter by adaptively testing the σ that meets the threshold condition; performs a fast Fourier transform (FFT) on the collected image to obtain a vignetting background in the frequency domain according to Gaussian filtering; extracts the vignetting background in sequence to obtain a correction lookup table, and then inverse Fourier transforms (IFT) are performed to obtain a corrected image by subtracting the correction lookup table from the original image; the corrected image is transferred to the frequency domain to iteratively update the amplitude to finally reconstruct a high-resolution image with uniform intensity. It has been verified that the method of removing the vignetting effect and then reconstructing the image by adaptive Gaussian filtering of the present invention can effectively avoid the influence of the vignetting effect on the quality of the stacked reconstructed image, improve the uneven intensity phenomenon of the reconstructed image, and also improve the image resolution. (IV) Description of the accompanying drawings
[0032] Figure 1 1. It is a schematic diagram of an algorithm flow of a Fourier stack microscopy imaging vignetting correction method based on adaptive Gaussian filtering according to the present invention;
[0033] Figure 2 This is a schematic diagram of the result of the vignetting correction reconstructed image of the present invention, and the color box diagram comparison shows that the image information is retained after filtering; (V) Specific implementation methods
[0034] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be understood as limiting the present invention.
[0035] See also Figure 1 The present invention provides a method for correcting vignetting in Fourier stack microscopy imaging based on adaptive Gaussian filtering, comprising the following steps:
[0036] S1: Set up a digital microscope optical path to collect the sample to be tested, record the multi-angle intensity map of the sample to be tested, and perform correction and reconstruction based on the intensity map;
[0037] S2: Set a threshold value that highly retains image information based on the standard deviation and mean of the current image, and obtain the optimal Gaussian parameter by testing σ that meets the threshold condition;
[0038] S3: The images are sequentially converted to the frequency domain by FFT and Gaussian filtering is used to obtain the vignetting frequency domain information, and the information is normalized and accumulated to obtain the correction lookup table;
[0039] S4: Subtract the IFT correction lookup table from the original image to obtain the corrected image;
[0040] S5: After correction, take the intermediate illumination as the target image and add the initial phase to the frequency domain. The phase remains unchanged. Use the image under the corresponding illumination angle to update the amplitude part to obtain the updated image light field.
[0041] S6: Perform Fourier transform on the updated image light field to obtain an updated spectrum, and add the updated spectrum components within the corresponding sub-aperture to the initial image spectrum;
[0042] S7: Repeat steps S5 to S6 until convergence, that is, minimizing the amplitude of the model-predicted reconstructed image and the true measured image.
[0043] In step S1, a Fourier stack microscopy system optical path based on flat-panel LED array illumination is used to acquire images at multiple angles and record them with a camera.
[0044] In step S2, by analyzing the standard deviation and mean of the image, the intensity range of the vignetting effect can be determined. The threshold setting filters out the areas most affected by vignetting and retains useful image information, thereby achieving adaptive adjustment and selecting the optimal Gaussian parameters, which can better match the spatial distribution characteristics of vignetting. The Gaussian filter modeling is expressed as:
[0045] G i (x,y)=exp(-((x i -u) 2 +(y i -v) 2 ) / 2*σ 2 )
[0046] Where (u,v) represents the center of the image, (x i ,y i ) is the pixel coordinate on the i-th image, σ is the standard deviation, which controls the width of the Gaussian function. The optimal matching parameters within the range are obtained through testing and comparison.
[0047] In step S3, MATLAB reads the captured image, performs FFT to obtain frequency domain information, applies Gaussian filtering to obtain vignetting information, normalizes it, and accumulates it to generate a correction lookup table (LUT). It can be expressed as:
[0048]
[0049] Among them I i Indicates the i-th image read, F(I i ) indicates a Fourier transform. The above indicates that after the image is Gaussian filtered, an inverse Fourier transform is performed to obtain the average to obtain the final LUT.
[0050] In step S4, the LUT obtained above is applied to perform correction processing on the multi-angle vignetting images in the spatial domain.
[0051] In step S5, the corrected image is reconstructed. During the reconstruction process, the image is transferred to the frequency domain and the phase is estimated by updating the amplitude, which is expressed as:
[0052]
[0053] in represents the updated target complex amplitude distribution corresponding to the LED in the mth row and nth column, is the low-resolution image captured corresponding to the LED unit in the mth row and nth column. The superscript u represents the updated target spectrum and target complex amplitude, and c represents the captured image. It is called the initial target complex amplitude.
[0054] In step S6, the spectrum of the updated target complex amplitude image is obtained using Fourier transform, and this low-resolution spectrum is used to update the spectral components within the corresponding sub-aperture in the object's high-resolution spectrum. The update formula is as follows:
[0055]
[0056] in The updated aperture spectrum is obtained by subtracting the initial spectrum from the updated target complex amplitude distribution spectrum corresponding to the LED in the mth row and nth column.
[0057] In step S7, the reconstructed image is repeatedly iterated. The following is the error of image reconstruction after a single iteration. The error is accumulated in each iteration, and then the designed error threshold is compared to determine whether the iteration has converged. The convergence formula is:
[0058]
[0059] Among them, Loss n (O,P) represents the loss function of the nth iteration, where O is the complex amplitude of the sample, I n It represents the low-resolution image captured at the n-th illumination angle in the n-th iteration, and the reconstructed image is obtained after the iteration ends.
[0060] The above disclosure is only a preferred embodiment of the present invention, and certainly cannot be used to limit the scope of the rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiment and equivalent changes made in accordance with the claims of the present invention are still within the scope of the invention.
Claims
1. A vignetting correction method for Fourier stack microscopy based on adaptive Gaussian filtering, characterized in that: The method comprises the following steps: S1: Set up a digital microscope optical path to collect the sample to be tested, record the multi-angle intensity map of the sample to be tested, and perform correction and reconstruction based on the intensity map; S2: Set a threshold value that highly retains image information based on the standard deviation and mean of the current image, and obtain the optimal Gaussian parameter by testing σ that meets the threshold condition; S3: The images are sequentially converted to the frequency domain by FFT and Gaussian filtering is used to obtain the vignetting frequency domain information, and the information is normalized and accumulated to obtain the correction lookup table; S4: Subtract the IFT correction lookup table from the original image to obtain the corrected image; S5: After correction, take the intermediate illumination as the target image and add the initial phase to the frequency domain. The phase remains unchanged. Use the image under the corresponding illumination angle to update the amplitude part to obtain the updated image light field. S6: Perform Fourier transform on the updated image light field to obtain an updated spectrum, and add the updated spectrum components within the corresponding sub-aperture to the initial image spectrum; S7: Repeat steps S5 to S6 until convergence, that is, minimizing the amplitude of the model-predicted reconstructed image and the true measured image to obtain the reconstructed image.
2. The method for vignetting correction of Fourier stack microscopy imaging based on adaptive Gaussian filtering according to claim 1, characterized in that: In S1, the digital microscope optical path is a Fourier stacked microscope imaging system optical path based on flat-panel LED array illumination.
3. The method for vignetting correction of Fourier stack microscopy imaging based on adaptive Gaussian filtering according to claim 1, characterized in that: In S2, by analyzing the standard deviation and mean of the image, the intensity range of the vignetting effect can be determined. The threshold setting filters out the areas most affected by vignetting and retains useful image information, thereby achieving adaptive adjustment and selecting the optimal Gaussian parameters, which can better match the spatial distribution characteristics of vignetting. The Gaussian filter modeling is expressed as: G i (x,y)=exp(-((x i -u) 2 +(y i -v) 2 ) / 2*σ 2 ) Where (u,v) represents the center of the image, (x i ,y i ) is the pixel coordinate on the i-th image, σ is the standard deviation, which controls the width of the Gaussian function. The optimal matching parameters within the range are obtained through testing and comparison.
4. The method for vignetting correction of Fourier stack microscopy imaging based on adaptive Gaussian filtering according to claim 1, characterized in that: In S3, MATLAB reads the collected image, performs FFT to obtain frequency domain information, applies Gaussian filtering to obtain vignetting information, normalizes it, and accumulates it to generate a correction lookup table (LUT). It can be expressed as: Among them I i Indicates the i-th image read, F(I i ) indicates a Fourier transform. The above indicates that after the image is Gaussian filtered, an inverse Fourier transform is performed to obtain the average to obtain the final LUT.
5. The method for vignetting correction of Fourier stack microscopy imaging based on adaptive Gaussian filtering according to claim 1, characterized in that: In S4, the LUT obtained above is used to perform correction processing on the multi-angle vignetting images in the spatial domain.
6. The method for Fourier stack microscopy imaging vignetting correction based on adaptive Gaussian filtering according to claim 1, characterized in that: In S5, the corrected image is reconstructed. During the reconstruction process, the image is transferred to the frequency domain and the phase is estimated by updating the amplitude, which is expressed as: represents the updated target complex amplitude distribution corresponding to the LED in the mth row and nth column, is the low-resolution image captured corresponding to the LED unit in the mth row and nth column. The superscript u represents the updated target spectrum and target complex amplitude, and c represents the captured image. It is called the initial target complex amplitude.
7. The method for Fourier stack microscopy imaging vignetting correction based on adaptive Gaussian filtering according to claim 1, characterized in that: In S6, the spectrum of the updated target complex amplitude image is obtained by Fourier transform, and this low-resolution spectrum is used to update the spectrum components within the corresponding sub-aperture in the high-resolution spectrum of the object. The update formula is as follows: in The updated aperture spectrum is obtained by subtracting the initial spectrum from the updated target complex amplitude distribution spectrum corresponding to the LED in the mth row and nth column.
8. The method for vignetting correction of Fourier stack microscopy imaging based on adaptive Gaussian filtering according to claim 1, characterized in that: In S7, the reconstructed image is repeatedly iterated. The following is the error of the image reconstruction after a single iteration. The error is accumulated in each iteration, and then the designed error threshold is compared to determine whether the iteration is converged. The convergence formula is: Among them, Loss n (O,P) represents the loss function of the nth iteration, where O is the complex amplitude of the sample, I n represents the low-resolution image captured at the n-th illumination angle in the n-th iteration.