A physical-based fourier-on-lens microscope light source pose correction method
By using physical models and optimization algorithms to correct light source pose deviations in Fourier layered microscopy, the image quality problem caused by light source position deviations was solved, achieving accurate reconstruction of high-resolution images and improving system robustness.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUILIN UNIV OF ELECTRONIC TECH
- Filing Date
- 2026-04-21
- Publication Date
- 2026-06-05
AI Technical Summary
In Fourier layered microscopy, the deviation of the light source position introduces phase error and amplitude error, which leads to artifacts and noise in the reconstructed image, affecting the image clarity and accuracy. Furthermore, algorithm-based correction methods are prone to getting trapped in local optima and are difficult to achieve globally optimal results.
By constructing a Fourier stacked microscopic optical path, low-resolution images of an empty glass slide are recorded. The center and radius of the boundary circles of the bright and dark fields are extracted using a physical model. Combined with an optimization algorithm, six pose parameters of the LED array are estimated, and the pose deviation of the light source is corrected to ensure the accuracy and uniformity of the lighting effect.
Precisely correcting the light source pose improves the quality of reconstructed high-resolution images, enhances the system's robustness to light source position deviations, reduces artifacts and noise, and improves image clarity and accuracy.
Smart Images
Figure CN122151330A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of Fourier layered microscopy, specifically relating to a physics-based method for correcting the pose of a Fourier layered microscopy light source. Background Technology
[0002] Fourier transform microscopy (FPM) is a novel quantitative phase imaging technique, initially developed by Zheng Guoan et al. in 2013. FPM integrates the concepts of phase retrieval and coherent synthetic aperture techniques. Compared to conventional microscopy imaging methods, this technique overcomes the resolution limitations set by the numerical aperture of the objective lens while maintaining the same field of view, thus achieving a significant improvement in image resolution. This has brought innovation to the field of microscopy.
[0003] FPM (Focused Permeation Microscopy) combines the concepts of phase retrieval and coherent synthetic aperture, solving the problem of resolution versus field of view constraints in traditional microscopy. It can acquire billion-pixel-level images without mechanical scanning and has been successfully applied in fields such as digital pathology in recent years. In Fourier transform microscopy, light source position deviation introduces additional phase and amplitude errors, leading to artifacts and noise in the reconstructed image, affecting image clarity and accuracy. LED position deviation reduces system robustness, making the system more sensitive to other errors (such as noise and optical distortion). Compensating for LED position deviation may require frequent adjustments to system parameters, increasing the difficulty of system adjustment and maintenance costs, and affecting the quality of the original image data. Therefore, correcting light source pose deviation is an indispensable part of improving the quality of Fourier transform imaging. Light source position correction methods can be divided into two categories: algorithm-based and physics-based methods. These two methods can also be integrated with deep learning-based FP (Focused Permeation) architectures to minimize alignment errors.
[0004] Algorithm-based methods involve solving two inverse problems: FPM reconstruction and system parameter calibration. It typically re-evaluates the differences between the captured and updated LR images at each iteration. In this regard, for example, a Global Position Misalignment (GPM) model has been introduced to characterize the illumination wave vector. Alternatively, a system calibration procedure called SC-FPM has been proposed to iteratively optimize the sub-aperture position while integrating a simulated annealing (SA) algorithm, an LED intensity correction module, and an adaptive step-size strategy into the calibration procedure. Although algorithm-based methods eliminate the need for prior knowledge and real-world testing, LR intensity measurements still contain a mixture of information from the sample and various errors. These methods are mostly based on GPM models; however, their effectiveness is limited by the optimization algorithms, leading to a tendency to obtain local optima rather than global optima.
[0005] Physics-based methods aim to separate system parameter calibration from the reconstruction process and locate precise sub-aperture positions using physical models. For example, the bright-field (BF) illumination angle can be located using the autocorrelation spectrum of the LR intensity image. BF alignment can be achieved using a bright-dark-field (BDF) transition image acquired without sample placement. Alternatively, misalignment parameters can be inferred by analyzing the offset BF position of defocused objects. The first method uses an additional 2×2 digital camera adapter to reduce the sampling aperture size, ensuring all BF Fourier spectra are within the LR image size range. The second method requires careful adjustment of the distance from the illumination source to the sample to ensure the presence of a BDF transition image. The defocusing strategy even requires a high-precision focus knob. Furthermore, the calibration process is primarily performed in the BF region, excluding the dark-field (DF) region. While information is more easily extracted from the BF region due to its higher intensity, the positional parameters in the BF do not accurately represent the true situation in the DF. Summary of the Invention
[0006] This invention proposes a physics-based Fourier layered microscopic light source pose correction method. It aims to address the problems caused by mechanical assembly errors, optical system errors, and electrical control errors in the system. These errors lead to inaccurate stitching of the sub-spectrums of low-resolution images in the Fourier domain, thereby reducing the quality of the final reconstructed high-resolution image.
[0007] To achieve the above objectives, this invention provides a physics-based Fourier layered microscopic light source pose correction method, the method comprising the following steps:
[0008] S1: Construct a Fourier stacked microscopic optical path to sample the empty slide under test and record a low-resolution image of the empty slide under test.
[0009] S2: Construct a Fourier stacked microscopic optical path to sample the empty slide under test and record a low-resolution image of the empty slide under test.
[0010] S3: Utilize the constructed physical model and extract the center coordinates of the boundary between the light and dark fields of these transition images.
[0011] S4: Extract the center and radius of the light and dark field boundaries from the captured transition image. Using the extracted center and radius, combined with the constructed physical model, use an optimization algorithm to estimate the six pose parameters of the LED array.
[0012] S5: The obtained six pose parameters are used to correct the light source pose deviation of the Fourier stacked microscopy system to improve the quality of the reconstructed high-resolution image.
[0013] In S1, the Fourier stacked microscopic optical path is a 4f system consisting of an array of LED light sources, a microscope objective, a tube lens, and an imaging camera.
[0014] In step S2, the relationship between the boundary information of the bright and dark fields in the captured image and the full pose parameters is as follows:
[0015]
[0016] in Represents the full pose parameters of the light source, ( , () represents the center coordinates of the boundary between the light and dark fields estimated using the constructed physical model. , () represents the center coordinates of the boundary between the light and dark fields of the actual captured image.
[0017] In S3, the transition image refers to a low-resolution image with bright and dark field boundaries.
[0018] In S3, the formula for calculating the center coordinates is:
[0019]
[0020]
[0021] in( , () represents the center coordinates of the LED in the m-th row and n-th column, and h represents the distance between the light source and the sample. This represents the distance between the entrance pupil and the sample. , , () represents the spatial coordinates of the LED in the m-th row and n-th column.
[0022] In step S4, due to the vignetting effect affecting the boundaries of the bright and dark fields in the actual captured transition image, the boundaries after binarization and boundary finding are not perfect arcs but rather jagged, making it extremely difficult to find the center and radius based on the arc length. Therefore, the transition image first needs to be denoised. After denoising, the boundary of the bright and dark fields in the transition image is found through boundary finding, and each pixel is converted into a two-dimensional point cloud to extract edge points and feature points. The feature points are used to initially align the boundaries, and the Iterative Closest Point (ICP) algorithm is used to minimize the distance between the feature points and the target circle. Finally, the least squares method is used to fit a circle based on the optimized feature points. This allows for a more accurate determination of the center coordinates and radius.
[0023] In step S4, an optimization algorithm is used to estimate the six pose parameters of the LED array. The optimization formula is as follows:
[0024]
[0025] In step S5, the formula for correcting the pose deviation of the light source is:
[0026]
[0027]
[0028]
[0029]
[0030] in( () represents the coordinates in the spectrum. R represents the coordinates in the spatial domain, and R represents the rotation matrix.
[0031] The beneficial effects of this invention are as follows:
[0032] This invention proposes a physics-based Fourier layered microscopic light source pose correction method. The method first constructs a physical model of the bright-to-dark field boundary of the captured image using the microscope. A set of low-resolution images of an empty slide are acquired, and transition images with bright-to-dark boundaries are selected. The bright-to-dark field boundary information of the transition images is compared with the unbiased transition images generated by the model to calculate the full pose parameters of the misaligned LED array (including distance from the sample, two orthogonal lateral displacements, in-plane rotation angles, and two tilt angles). This allows for precise correction of the illumination angle, ensuring the accuracy and uniformity of the illumination effect. Experimental verification shows that this invention achieves excellent correction results in practical applications, enhancing the robustness of the Fourier layered reconstruction system to light source position deviations. Attached Figure Description
[0033] Figure 1 This is a schematic diagram of a physics-based Fourier layered microscopic light source pose correction method according to the present invention.
[0034] Figure 2 It is an algorithmic fit to the boundary between bright and dark fields in a real-world transition image.
[0035] Figure 3 The left figure is a schematic diagram of the reconstruction result before the light source pose correction according to the present invention. Figure 3 The right figure is a schematic diagram of the corrected reconstruction result. Detailed Implementation
[0036] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.
[0037] A physics-based Fourier layered microscopy light source pose correction method includes the following steps:
[0038] S1: Construct a Fourier stacked microscopic optical path to sample the empty slide under test and record a low-resolution image of the empty slide under test.
[0039] S2: Construct a Fourier stacked microscopic optical path to sample the empty slide under test and record a low-resolution image of the empty slide under test.
[0040] S3: Utilize the constructed physical model and extract the center coordinates of the boundary between the light and dark fields of these transition images.
[0041] S4: Extract the center and radius of the light and dark field boundaries from the captured transition image. Using the extracted center and radius, combined with the constructed physical model, use an optimization algorithm to estimate the six pose parameters of the LED array.
[0042] S5: The obtained six pose parameters are used to correct the light source pose deviation of the Fourier stacked microscopy system to improve the quality of the reconstructed high-resolution image.
[0043] In S1, the Fourier stacked microscopic optical path is a 4f system consisting of an array of LED light sources, a microscope objective, a tube lens, and an imaging camera.
[0044] In step S2, the relationship between the boundary information of the bright and dark fields in the captured image and the full pose parameters is as follows:
[0045]
[0046] in Represents the full pose parameters of the light source, ( , () represents the center coordinates of the boundary between the light and dark fields estimated using the constructed physical model. , () represents the center coordinates of the boundary between the light and dark fields of the actual captured image.
[0047] In S3, the transition image refers to a low-resolution image with bright and dark field boundaries.
[0048] In S3, the formula for calculating the center coordinates is:
[0049]
[0050]
[0051] in( , () represents the center coordinates of the LED in the m-th row and n-th column, and h represents the distance between the light source and the sample. This represents the distance between the entrance pupil and the sample. , , () represents the spatial coordinates of the LED in the m-th row and n-th column.
[0052] In step S4, due to the vignetting effect affecting the boundaries of the bright and dark fields in the actual captured transition image, the boundaries after binarization and boundary finding are not perfect arcs but rather jagged, making it extremely difficult to find the center and radius based on the arc length. Therefore, the transition image first needs to be denoised. After denoising, the boundary of the bright and dark fields in the transition image is found through boundary finding, and each pixel is converted into a two-dimensional point cloud to extract edge points and feature points. The feature points are used to initially align the boundaries, and the Iterative Closest Point (ICP) algorithm is used to minimize the distance between the feature points and the target circle. Finally, the least squares method is used to fit a circle based on the optimized feature points. This allows for a more accurate determination of the center coordinates and radius.
[0053] In step S4, an optimization algorithm is used to estimate the six pose parameters of the LED array. The optimization formula is as follows:
[0054]
[0055] In step S5, the formula for correcting the pose deviation of the light source is:
[0056]
[0057]
[0058]
[0059]
[0060] in( () represents the coordinates in the spectrum. R represents the coordinates in the spatial domain, and R represents the rotation matrix.
[0061] The above description discloses only one preferred embodiment of the present invention, and should not be construed as limiting the scope of the present invention. Those skilled in the art will understand that all or part of the processes of the above embodiments can be implemented, and equivalent changes made in accordance with the claims of the present invention are still within the scope of the invention.
Claims
1. A physics-based Fourier layered microscopic light source pose correction method, characterized in that, The method includes the following steps: S1: Construct a Fourier stacked microscopic optical path to sample the empty slide under test and record a low-resolution image of the empty slide under test. S2: A physics-based model was constructed that connects the light and dark field boundary information in the captured image with the full pose parameters. S3: Utilize the constructed physical model and extract the center coordinates of the boundary between the light and dark fields of these transition images. S4: Extract the center and radius of the light and dark field boundaries from the captured transition image. Using the extracted center and radius, combined with the constructed physical model, use an optimization algorithm to estimate the six pose parameters of the LED array. S5: The obtained six pose parameters are used to correct the light source pose deviation of the Fourier stacked microscopy system to improve the quality of the reconstructed high-resolution image.
2. The method for pose correction of a physics-based Fourier layered microscopic light source according to claim 1, characterized in that: In S1, the Fourier stacked microscopic optical path is a 4f system consisting of an array of LED light sources, a microscope objective, a tube lens, and an imaging camera.
3. The method for correcting the pose of a physical Fourier layered microscopic light source according to claim 1, characterized in that: In S1, a set of low-resolution images is obtained by illuminating LEDs at different locations.
4. The method for pose correction of a physics-based Fourier layered microscopic light source according to claim 1, characterized in that: In step S2, the relationship between the boundary information of the bright and dark fields in the captured image and the full pose parameters is as follows: in Represents the full pose parameters of the light source, ( , () represents the center coordinates of the boundary between the light and dark fields estimated using the constructed physical model. , () represents the center coordinates of the boundary between the light and dark fields of the actual captured image.
5. The method for pose correction of a physics-based Fourier layered microscopic light source according to claim 1, characterized in that: In S3, the transition image refers to a low-resolution image with bright and dark field boundaries.
6. The method for pose correction of a physics-based Fourier layered microscopic light source according to claim 1, characterized in that: In S3, the formula for calculating the center coordinates is: in( , () represents the center coordinates of the LED in the m-th row and n-th column, and h represents the distance between the light source and the sample. This represents the distance between the entrance pupil and the sample. , , () represents the spatial coordinates of the LED in the m-th row and n-th column.
7. The method for pose correction of a physics-based Fourier layered microscopic light source according to claim 1, characterized in that: In step S4, due to the vignetting effect affecting the boundaries of the bright and dark fields in the actual captured transition image, the boundaries after binarization and boundary finding are not perfect arcs but rather jagged, making it extremely difficult to find the center and radius based on the arc length. Therefore, the transition image first needs to be denoised. After denoising, the boundary of the bright and dark fields in the transition image is found through boundary finding, and each pixel is converted into a two-dimensional point cloud to extract edge points and feature points. The feature points are used to initially align the boundaries, and the Iterative Closest Point (ICP) algorithm is used to minimize the distance between the feature points and the target circle. Finally, the least squares method is used to fit a circle based on the optimized feature points. This allows for a more accurate determination of the center coordinates and radius.
8. The method for pose correction of a physics-based Fourier layered microscopic light source according to claim 1, characterized in that: In step S4, an optimization algorithm is used to estimate the six pose parameters of the LED array. The optimization formula is as follows: 。 9. The method for pose correction of a physics-based Fourier layered microscopic light source according to claim 1, characterized in that: In step S5, the formula for correcting the pose deviation of the light source is: in( () represents the coordinates in the spectrum. R represents the coordinates in the spatial domain, and R represents the rotation matrix.