Method and system for realizing vignetting correction of Fourier laminated imaging by adopting weighted fusion strategy
By correcting the vignetting effect in Fourier layered imaging using a weighted fusion strategy, the problem of uneven image brightness was solved, high-quality image reconstruction was achieved, and imaging resolution and contrast were improved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HANGZHOU DIANZI UNIV
- Filing Date
- 2026-01-07
- Publication Date
- 2026-04-10
AI Technical Summary
Existing Fourier layer imaging techniques suffer from vignetting during the imaging process, resulting in uneven brightness during image reconstruction, which affects resolution and contrast. Existing correction methods rely on additional calibration or are computationally complex, limiting their versatility.
A weighted fusion strategy is adopted. By acquiring images under no-light conditions and performing dark field correction, a normalized illumination intensity map is generated and subjected to exponential transformation. A two-dimensional Gaussian filter is used to smooth the weight map, and a safety threshold and attenuation coefficient are introduced to achieve weighted fusion correction of the image.
It effectively eliminates the uneven brightness caused by vignetting, preserves target information, suppresses background noise, provides high-quality image input, and improves imaging performance.
Smart Images
Figure CN121842525A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of optoelectronic imaging and computer image processing technology, and specifically relates to a method and system for vignetting correction of Fourier layered imaging using a weighted fusion strategy. Background Technology
[0002] Fourier Ptychographic Microscopy (FPM) is a computational imaging technique based on frequency domain stitching. It acquires multiple low-resolution intensity images under illumination at different angles and reconstructs them in the frequency domain using phase retrieval algorithms. Compared to traditional microscopy, FPM overcomes the inherent contradiction between resolution and field of view in traditional optical microscopes, achieving high-throughput and high-precision imaging. It provides a powerful analytical tool for fields such as biomedicine, materials science, and industrial inspection.
[0003] However, in the actual acquisition and reconstruction process of FPM, the system often exhibits a vignetting effect. The vignetting effect is a brightness attenuation phenomenon caused by imperfect matching between the illumination optical path and the numerical aperture of the imaging system, uneven transmittance at the lens edges, or off-axis illumination of the optical system. This manifests as lower brightness at the image edges compared to the central area. This uneven illumination leads to an imbalance in the amplitude of each sub-spectrum, introducing spectral stitching errors, reducing the contrast and resolution of the reconstructed image, and causing instability in the phase retrieval algorithm.
[0004] Current common methods for vignetting correction mainly include preprocessing methods based on flat-field correction, empirical modeling methods using background fitting, and iterative optimization methods that introduce brightness normalization during reconstruction. The first two methods rely on additional calibration experiments or prior models and have limited versatility; the latter is computationally intensive and complex to implement. Summary of the Invention
[0005] This invention proposes a method and system for vignetting correction in Fourier layered imaging using a weighted fusion strategy, aiming to solve the problem of uneven intensity in FPM reconstructed images caused by vignetting phenomena where the center of the sub-aperture image is bright and the surrounding area is dark.
[0006] To achieve the above objectives, the present invention provides a method for vignetting correction in Fourier layered imaging using a weighted fusion strategy, the method comprising the following steps: S1: Acquire several dark-field and flat-field images under no-light conditions. and sample images Median filtering is performed on the dark field image to obtain the dark noise template. Then, dark field correction was performed on the flat field image and the sample image respectively.
[0007] S2: Extract the global maximum gray value of the flat-field image after dark field correction. The entire flat-field image is normalized to the standard range of 0 to 1 to generate a normalized illumination intensity map. .
[0008] S3: Perform an exponential transformation on the lighting intensity map to generate an initial weight map. .
[0009] S4: Use a two-dimensional Gaussian filter to spatially smooth the initial weight map, achieving a smooth transition from the central high-weight region to the edge low-weight region, thus generating a smooth weight map. .
[0010] S5: Introduces a safety threshold mechanism, performing division correction only on pixels whose grayscale values exceed a threshold in the flat-field image. The sample image is divided by the flat-field image to generate the foreground correction image. .
[0011] S6: Divide the sample image by the maximum gray value of the flat field to achieve global normalization and generate a background-corrected image. .
[0012] S7: Based on smooth weight graph The foreground precisely corrected image and the background globally normalized image are weighted and fused; an attenuation coefficient is introduced to further suppress the background and achieve homogenization. This achieves the correction of the vignetting effect.
[0013] In another aspect, the present invention provides a vignetting correction system for Fourier layered imaging using a weighted fusion strategy, comprising the following modules: The dark field correction module is used to acquire several dark field images, flat field images and sample images under no-light conditions, and to perform dark field correction on the flat field images and sample images respectively. The initial weight map module is used to extract the global maximum gray value of the flat field image after dark field correction, normalize the entire flat field image, generate an illumination intensity map, and perform an exponential transformation on the illumination intensity map to generate the initial weight map. The Smoothing Weight Graph module is used to perform spatial domain smoothing on the initial weight graph using a two-dimensional Gaussian filter to generate a smooth weight graph. The foreground correction image module is used to introduce a safety threshold mechanism, which performs division correction on pixels whose gray values in the flat field image exceed the threshold, and divides the sample image by the flat field image to generate the foreground correction image. The background correction image module is used to divide the sample image by the maximum gray value of the flat field to achieve global normalization and generate a background correction image. The vignetting correction output module is used to perform weighted fusion of the foreground correction image and the background correction image based on a smooth weight map to correct the vignetting effect.
[0014] This invention presents a method for vignetting correction in Fourier layered imaging using a weighted fusion strategy. Based on this strategy, the invention achieves adaptive processing by applying precise correction to high-weight regions (image center) and attenuation suppression to low-weight regions (image edges). This effectively eliminates brightness unevenness caused by vignetting, preserves target information while suppressing background noise, and achieves accurate and efficient vignetting correction. This provides high-quality input for subsequent image reconstruction and improves the overall performance of Fourier layered microscopy. Attached Figure Description
[0015] Figure 1 This is a flowchart illustrating the method for achieving Fourier layered imaging vignetting correction using the weighted fusion strategy of the present invention. Figure 2 This is a comparative diagram of low-resolution images after vignetting correction. Detailed Implementation
[0016] like Figure 1 As shown, in one aspect, the present invention provides a method for vignetting correction in Fourier layered imaging using a weighted fusion strategy. The implementation method of the present invention includes the following steps: S1: Data collection and loading, and dark field correction; First, several dark-field images were acquired under no-light conditions, followed by 225 sample images and corresponding flat-field images. For the dark-field images, a dark noise template was obtained through median filtering. This effectively suppresses the effects of random noise.
[0017] For both ordinary and sample images, subtraction correction, i.e., dark field correction, is performed: The max function ensures that the corrected pixel value is non-negative.
[0018] S2: Extract the maximum grayscale value of the entire image and generate a normalized lighting map; Lighting intensity distribution map The expression is: This image records a spatially non-uniform illumination distribution. The value of each pixel in the image represents the actual light intensity received at that location, with its maximum value being... This typically occurs near the center of the optical axis and represents the peak illumination intensity the system can provide. Specifically, the light intensity at any location in the image will not exceed the light intensity at the center of the optical axis. The value ranges from 0 to 1. At the center of maximum light intensity, this ratio is close to 1; at the edge of light intensity decay, this ratio approaches 0.
[0019] S3: Perform an exponential transformation on the lighting intensity map to generate an initial weight map; Initial weight graph The expression is: Directly using normalized illumination maps as weights is often too flat and makes it difficult to clearly distinguish between "high signal-to-noise ratio central regions" and "low signal-to-noise ratio edge regions." By introducing a significance index... An exponential transformation is performed to adjust the steepness of the weight distribution, ensuring that the algorithm can switch to the background suppression strategy more quickly in edge areas with extremely low illumination, thereby avoiding the false amplification of noise. The smaller the value, the more pronounced the edge decay. The value can be adjusted based on the actual effect. The value range is 0.5 to 1.5.
[0020] S4: Spatial domain smoothing of the weight map is performed using a two-dimensional Gaussian filter; Generate a smooth weighted graph The expression is: The initial weight map obtained from S3 is calculated from the actual acquired flat-field image. Due to sensor characteristics, it often contains high-frequency random noise. If the noisy weight map is used directly for fusion, granular artifacts will be introduced into the final image. Gaussian smoothing eliminates pixel-level abrupt changes and noise interference in the weight map, ensuring the continuity of the transition from the central "precise correction area" to the edge "background suppression area," and avoiding obvious artificial stitching marks or false contours at the image fusion point.
[0021] S5: Generate foreground correction image Based on the correction idea: The correction formula can be derived as follows: In the edge region where vignetting is severe, flat-field images The grayscale values are extremely low, and direct division will amplify tiny perturbations in the denominator into huge numerical noise. Therefore, although this step achieves the most realistic contrast restoration in the central region, it will produce severe salt-and-pepper noise in the edge regions, which needs to be avoided through subsequent weighted blending steps.
[0022] S6: Global normalization, generating a background-corrected image; Background correction image The expression is: Unlike foreground correction, this step abandons pixel-by-pixel physical intensity recovery and instead uses global normalization. This strategy avoids the noise explosion problem caused by "dividing by a small value" in S5.
[0023] S7: Perform weighted fusion of the foreground precisely corrected image and the background globally normalized image.
[0024] The purpose of foreground-background contrast control is to selectively highlight the foreground while suppressing background clutter and noise. The expression for weighted fusion is: This step utilizes a smoothed weight map. As a spatial selector, in well-lit centers, the weight is close to 1, and the system primarily relies on high-precision foreground correction images. At poorly lit edges, the weights approach 0, and the system automatically switches to a low-noise background-corrected image. And supplemented by attenuation coefficient Further suppression of background clutter. This strategy resolves the contradiction in traditional methods between "introducing edge noise to remove vignetting" and "failing to remove vignetting to preserve noise." Final corrected image. It maintains high contrast and physical realism in the central region while achieving a smooth and noise-free transition in the edge region, providing high-quality, high-signal-to-noise-ratio spectral input for the subsequent FPM phase retrieval algorithm, thereby significantly improving the reconstruction success rate.
[0025] See Figure 2 The left side shows the initial low-resolution sample image acquired. Due to the numerical aperture mismatch of the microscope system and the decrease in lens transmittance, the image exhibits a significant vignetting effect: the brightness in the central area of the image is normal, and the stripe details of the resolution plate can be clearly distinguished; as the field of view extends towards the edge, the brightness decreases sharply, causing the edge area to fall into darkness, making it difficult to distinguish the target information from the background.
[0026] Figure 2 The right side shows the corrected image processed by the weighted fusion strategy of this invention. It can be seen that the corrected image achieves uniform brightness distribution across the entire field of view. Edge stripes that were originally "invisible" due to insufficient lighting are clearly restored, and the overall image contrast is significantly improved.
[0027] In another aspect, the present invention provides a vignetting correction system for Fourier layered imaging using a weighted fusion strategy, comprising the following modules: The dark field correction module is used to acquire several dark field images, flat field images and sample images under no-light conditions, and to perform dark field correction on the flat field images and sample images respectively. The initial weight map module is used to extract the global maximum gray value of the flat field image after dark field correction, normalize the entire flat field image, generate an illumination intensity map, and perform an exponential transformation on the illumination intensity map to generate the initial weight map. The Smoothing Weight Graph module is used to perform spatial domain smoothing on the initial weight graph using a two-dimensional Gaussian filter to generate a smooth weight graph. The foreground correction image module is used to introduce a safety threshold mechanism, which performs division correction on pixels whose gray values in the flat field image exceed the threshold, and divides the sample image by the flat field image to generate the foreground correction image. The background correction image module is used to divide the sample image by the maximum gray value of the flat field to achieve global normalization and generate a background correction image. The vignetting correction output module is used to perform weighted fusion of the foreground correction image and the background correction image based on a smooth weight map to correct the vignetting effect.
[0028] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Those skilled in the art can understand that all or part of the processes of the above embodiments can be implemented. Any equivalent substitutions made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A Fourier ptychographic vignetting correction method using a weighted fusion strategy, characterized in that, The method comprises the following steps: S1: acquiring a plurality of dark field images, flat field images and sample images under no light condition, and performing dark field correction on the flat field images and the sample images respectively; S2: extracting a global maximum gray value of the dark field corrected flat field image, normalizing the flat field image, generating an illumination intensity map, performing exponential transformation on the illumination intensity map, and generating an initial weight map; S3: performing spatial domain smoothing on the initial weight map using a two-dimensional Gaussian filter to generate a smoothed weight map; S4: introducing a safety threshold mechanism, performing division correction on the pixels of the flat field image whose gray value exceeds the threshold, dividing the sample image by the flat field image to generate a foreground corrected image; S5: dividing the sample image by the maximum gray value of the flat field image to realize global normalization, and generating a background corrected image; S6: based on the smoothed weight map, the foreground corrected image and the background corrected image are weighted and fused to realize the correction of the vignetting effect.
2. The method of claim 1, wherein the weighted fusion strategy is implemented to correct the vignetting in the Fourier ptychographic imaging. The S1 is specifically implemented as: collecting a dark field image and a flat field image under no light condition and a sample image , performing median filtering on the dark field image to obtain a dark noise template , and then performing dark field correction on the flat field image and the sample image respectively to obtain and , and the process expression is: 。 3. The method of claim 2, wherein the weighted fusion strategy is implemented to correct the vignetting in the Fourier ptychographic imaging. The specific process of generating the illumination intensity map is as follows: extracting a global maximum grey value of the dark-field corrected flat-field image normalizing the entire flat-field image to generate a normalized illumination intensity map ; maximum gray value The expression is: Light intensity distribution diagram The expression is: representing the relative intensity of illumination at the pixel coordinate representing the pixel coordinate in the dark-field corrected flat-field image representing the pixel coordinate in the dark-field corrected flat-field image 4. The method for vignetting correction of Fourier layered imaging using a weighted fusion strategy according to claim 3, characterized in that, The specific process of generating the initial weight map is as follows: initial weight map The expression is: wherein is a saliency index used to control the steepness of the weight distribution.
5. The method of claim 4, wherein the weighted fusion strategy is implemented as a Fourier ptychographic method. The S3 is specifically implemented as follows: a Gaussian filter with a standard deviation of σ is used to smooth the initial weight map, and the expression is: To generate a smoothed weight map.
6. The method of claim 5, wherein the weighted fusion strategy is implemented to correct the vignetting in the Fourier ptychographic imaging. The S4 is specifically implemented as follows: Foreground corrected image The expression is: wherein, represents the pixel coordinate at the position in the dark-field corrected sample image, is a safety threshold.
7. The method of claim 6, wherein the weighted fusion strategy is implemented to correct the vignetting in the Fourier ptychographic imaging. The S5 is specifically implemented as follows: Background corrected image The expression is: 。 8. The method of claim 7, wherein the weighted fusion strategy is implemented to correct the vignetting in the Fourier ptychographic imaging. The S6 is specifically implemented as follows: According to the smoothing weight map The foreground corrected image and the background corrected image are fused by the following expression: where a is the background attenuation coefficient, is the final corrected image.
9. A system for Fourier ptychographic vignetting correction using a weighted fusion strategy, for implementing the method for Fourier ptychographic vignetting correction using a weighted fusion strategy according to any one of claims 1 to 8, characterized in that, The method comprises the following modules: A dark field correction module is configured to acquire a plurality of dark field images, flat field images and sample images under no light condition, and perform dark field correction on the flat field images and the sample images respectively; An initial weight map module is configured to extract a global maximum gray value of the dark field corrected flat field image, normalize the flat field image, generate an illumination intensity map, perform exponential transformation on the illumination intensity map, and generate an initial weight map; A smoothed weight map module is configured to perform spatial domain smoothing on the initial weight map using a two-dimensional Gaussian filter to generate a smoothed weight map; A foreground corrected image module is configured to introduce a safety threshold mechanism, perform division correction on the pixels of the flat field image whose gray value exceeds the threshold, divide the sample image by the flat field image to generate a foreground corrected image; A background corrected image module is configured to divide the sample image by the maximum gray value of the flat field image to realize global normalization and generate a background corrected image; A vignetting correction output module is configured to, based on the smoothed weight map, weight and fuse the foreground corrected image and the background corrected image to realize the correction of the vignetting effect.