Microscopic image vignetting correction method

Abstract

The invention provides a microscopic image vignetting correction method, which comprises the following steps of: firstly, converting an image to be processed into an HSV color mode, and decomposing the image into a hue channel image H, a saturation channel image S and a brightness channel image V; then, an optimization equation is constructed and solved, and a brightness imbalance factor matrix L and a dark field layer D are obtained; and finally, calculating a corrected image of the V-channel image, re-synthesizing the corrected image, the original H-channel image and the S-channel image into an HSV image, and converting the obtained HSV image into an RGB image to complete correction. According to the method, the optimization equation is constructed based on the imaging physical principle, the sparsity characteristic of the brightness imbalance factor matrix is utilized, and the brightness imbalance factor matrix is subjected to regularization constraint through the L1 norm, so that the brightness imbalance factor matrix of the vignetting image is solved and obtained, and then efficient vignetting correction is achieved.

Description

Technical Field

[0001] This invention relates to the field of microscopic optical imaging technology, and specifically to a method for correcting vignetting in microscopic images. Background Technology

[0002] Electron microscopy imaging equipment, as a core tool in industrial inspection and scientific research analysis, directly determines the accuracy of results in key applications such as PCB circuit defect identification and biological cell observation. A common problem in these devices is vignetting (commonly known as "dim screen"), which manifests as a significant decrease in brightness at the image edges. This phenomenon stems from three main physical mechanisms: first, the inherently lower sensitivity of the sensor edges to incident light; second, the mechanical obstruction caused by the light-shielding structure; and third, energy loss due to the off-axis light source following the fourth-power cosine decay law.

[0003] With the large-scale application of large-area CCD sensors, the vignetting phenomenon exhibits a non-linear intensification characteristic—not only is the edge brightness decay rate no longer uniform, but it is also accompanied by a significant decrease in the signal-to-noise ratio of image edge features. This directly leads to a significant reduction in the stability of automatic detection systems in judging edge defects, becoming a key bottleneck restricting the improvement of detection accuracy.

[0004] Traditional vignetting removal methods have significant limitations: these methods are mostly based on the calculation of the original image entropy value and assume that the geometric center of the image is the optical center, assuming that the brightness decreases uniformly from the center to the edges. However, in real-world imaging scenarios, the offset between the optical center and the geometric center is a common phenomenon, which greatly reduces the correction effect of traditional algorithms.

[0005] The existing invention patent CN113888419B addresses this issue by introducing an optical center search mechanism that minimizes image intensity entropy to handle vignetting, and by using a polynomial gain function to model the brightness distribution, thus improving the vignetting correction effect to some extent. However, this scheme still has two key drawbacks: first, the assumption of the optical center is too rigid; when the deviation between the vignetting center and the image center exceeds 15%, the parameter fitting of the polynomial gain function fails, and the correction effect deteriorates sharply; second, it does not consider the spatial non-uniformity of the actual vignetting distribution—the brightness attenuation of real vignetting images varies significantly in different directions, unlike the isotropic distribution assumed by traditional models. Furthermore, other existing technologies also struggle to meet the correction requirements of high-precision electron microscopy imaging. Summary of the Invention

[0006] The purpose of this invention is to provide a method for correcting vignetting in microscopic images, which is an efficient method for correcting vignetting that can adapt to off-center scenes and cope with non-uniform attenuation, so as to solve the image quality bottleneck of electron microscopy imaging equipment.

[0007] The present invention provides a method for correcting vignetting in microscopic images, comprising the following steps:

[0008] Step (1): Obtain the vignetting microscopic image to be processed, and convert the image from RGB color mode to HSV color mode microscopic image; decompose the HSV image into hue channel image H, saturation channel image S and luminance channel image V;

[0009] Step (2): Construct and solve the optimization equation to obtain the brightness imbalance factor matrix L and the dark field layer D;

[0010] The optimization equation is:

[0011]

[0012] in,

[0013] Denotes the L1 norm of a matrix. G represents the corrected image. f (·) represents the Gaussian low-pass filter function, G m (·) denotes the median filter function, D m represents the average value of the dark field layer matrix D, min(V) represents the minimum value of the luminance channel image matrix V, and ⊙ represents element-wise multiplication of the matrices. This indicates the addition of corresponding elements of the matrix. This represents matrix multiplication. To optimize the residual terms in the equation, and To optimize the regularization term in the equation, α1 and α2 are two regularization parameters;

[0014] Step (3): Calculate the corrected image of the V channel image. Correct the image The original H hue channel image and S saturation channel image are recombined to form an HSV image; the resulting HSV image is then converted to an RGB image to complete the correction.

[0015] As a further preferred option of the above scheme:

[0016] In the optimization equation, the residual term constructs a function F(·) that is the Shannon-Wiener entropy, and the formula is:

[0017]

[0018] in, Represents the corrected image First use G m (·) Median filtering operation, followed by G(·) Gaussian low-pass filtering function operation. This represents the result after median and Gaussian filtering.

[0019] As a further preferred option of the above scheme:

[0020] The formula China G f (·) is the Gaussian low-pass filter function, G f The standard deviation σ in (·) is 3.2; the median filter function G m The template size for (·) is 3x3. express The value of the i-th pixel Probability, N is Number of pixels

[0021] The beneficial effects of this invention are as follows: The microscopic image vignetting correction method provided in this application constructs an optimization equation based on imaging physics principles, fully utilizes the sparsity characteristics of the brightness imbalance factor matrix, and applies a regularization constraint to this matrix through the L1 norm to accurately solve the brightness imbalance factor matrix of the vignetting image; simultaneously, it introduces the Shannon-Wiener entropy, preprocessed by median filtering and Gaussian low-pass filtering, to apply a secondary regularization constraint to the image brightness difference, ultimately achieving efficient and accurate microscopic image vignetting correction. Specific beneficial effects are as follows:

[0022] For off-center scenes where the center of the light source deviates from the center of the image, and for the problem of non-uniform attenuation vignetting in the image, the correction effect of this invention is significantly superior to existing technologies. For typical vignetting microscopic images with significant differences in darkness at the four corners and a brighter upper central area, existing technologies still exhibit local brightness imbalances after correction, such as some corners not being effectively darkened and some corners being brighter than the central area, failing to achieve brightness balance across the entire image. However, after correction using this invention, the entire image visually presents a near-uniform brightness state, with the brightness of the originally darker corner areas being basically the same as that of the central area. This achieves an ideal correction effect where the brightness of the entire image tends to be consistent, except for the target line texture, thus effectively improving the vignetting correction quality for this type of microscopic image. Attached Figure Description

[0023] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0024] Figure 1 This is a flowchart of the method of the present invention;

[0025] Figure 2 This is a microscopic image of the vignetting to be corrected in a specific embodiment of the present invention;

[0026] Figure 3It is based on the invention patent CN113888419B in the comparative example. Figure 2 The corrected image obtained after processing;

[0027] Figure 4 It is based on the invention patent CN113888419B in the comparative example. Figure 2 The anti-diagonal image numerical curve of the corrected image obtained after processing and Figure 2 The brightness curve formed by pixels on the opposite diagonal;

[0028] Figure 5 It is based on the invention patent CN113888419B in the comparative example. Figure 2 The diagonal image numerical curve of the corrected image obtained after processing and Figure 2 The brightness curve formed by pixels on the diagonal line;

[0029] Figure 6 According to the method of the present invention Figure 2 The corrected image obtained after processing;

[0030] Figure 7 According to the method of the present invention Figure 2 The brightness imbalance factor matrix obtained after correction;

[0031] Figure 8 According to the method of the present invention Figure 2 The anti-diagonal image numerical curve of the corrected image obtained after processing and Figure 2 The brightness curve formed by pixels on the opposite diagonal;

[0032] Figure 9 According to the method of the present invention Figure 2 The diagonal image numerical curve of the corrected image obtained after processing and Figure 2 The brightness curve formed by pixels on the diagonal. Detailed Implementation

[0033] The present invention will be further illustrated below with reference to specific embodiments. These embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. The structural features of the present invention will now be described in detail with reference to the accompanying drawings.

[0034] Example

[0035] See Figure 1 This is a flowchart of the method in this embodiment. This embodiment provides a method for correcting vignetting in microscopic images, including the following steps:

[0036] Step (1): RGB to HSV color mode conversion

[0037] Obtain the vignetting micrograph to be processed, and convert the image from RGB color mode to HSV color mode. See the vignetting micrograph to be processed. Figure 2 ;

[0038] Step (2), HSV image decomposition

[0039] The HSV image in step (1) is decomposed into a hue channel image H, a saturation channel image S, and a luminance channel image V;

[0040] Step (3): Construct and solve the optimization equation to obtain the brightness imbalance factor matrix L and the dark field layer D.

[0041] Obtain the brightness imbalance factor matrix L and the dark field layer D of the brightness channel image V in step (2). See the image of the L matrix. Figure 3 Based on the principles of imaging physics, an optimization equation is constructed and solved as shown in formula (1) to obtain the brightness imbalance factor matrix L and the dark field layer D:

[0042]

[0043] in,

[0044] Denotes the L1 norm of a matrix. G represents the corrected image. f (·) represents the Gaussian low-pass filter function, G m (·) denotes the median filter function, D m represents the average value of the dark field layer matrix D, min(V) represents the minimum value of the luminance channel image matrix V, and ⊙ represents element-wise multiplication of the matrices. This indicates the addition of corresponding elements of the matrix. This represents matrix multiplication. To optimize the residual terms in the equation, and To optimize the regularization term in the equation, α1 and α2 are two regularization parameters;

[0045] Furthermore, in formula (1), the residual term constructs the function F(·), which is the Shannon-Wiener entropy, specifically:

[0046]

[0047] in, Represents the corrected image First use G m (·) Median filtering operation, followed by G(·) Gaussian low-pass filtering function operation. This represents the result after median and Gaussian filtering. The Gaussian low-pass filter function G... fIn (·), the standard deviation σ takes a value of 3.2, and the median filter function G m The template size for (·) is 3x3. express The value of the i-th pixel Probability, N is The number of pixels;

[0048] Step (4): Calculate the corrected image of the V channel image.

[0049] After obtaining the luminance imbalance factor matrix L of the luminance channel image V and the dark field layer D using step (3), the corrected image is calculated using the following formula.

[0050]

[0051] Step (5), HSV image synthesis

[0052] The corrected image of the V channel image obtained in step (4) The original H (hue) channel image and S (saturation) channel image are then recombined to form an HSV image.

[0053] Step (6): HSV to RGB color mode conversion

[0054] Convert the HSV image obtained in step (5) to an RGB image to complete the correction. See the image below for the corrected result. Figure 6 .

[0055] Comparative Example

[0056] In the embodiment, the original vignetting image (such as...) Figure 2 The image shown has typical characteristics: except for a few target lines and textures, the rest of the area is a single background; the image exhibits a distribution pattern where the four corners are darker with significant differences in the degree of vignetting (the lower left corner is the darkest), and the upper center area is brighter, which is a typical image where the center of the light source deviates from the center of the image. Based on this, the ideal correction effect should be: except for the target lines and textures, the brightness of the entire image should be uniform.

[0057] Using a prior art image vignetting removal method (publication number CN113888419B) as a comparison, the correction result after processing is as follows: Figure 3 As shown. By Figure 3 As can be seen, after correction by the comparative patent, the lower left corner of the image remains relatively dark, and the brightness of the upper left and upper right corners is higher than that of the central area; Figure 4 for Figure 3 The brightness curve formed by the pixels along the diagonal line from the top left corner to the bottom right corner. Figure 5 for Figure 3 The brightness curve formed by pixels along the diagonal line from the bottom left corner to the top right corner. (Combined with...) Figure 3 , Figure 4 and Figure 5 It can be seen that the correction process of the comparative patent did not achieve a near-balanced brightness across the entire image, indicating that the comparative patent's correction effect on this type of microscopic image is unsatisfactory.

[0058] After correction using the present invention, the correction result is as follows: Figure 7 As shown. By Figure 7 As can be seen, after correction by the method of this invention, the entire image presents a nearly uniform brightness distribution visually; Figure 8 for Figure 7 The brightness curve formed by the pixels along the diagonal line from the top left corner to the bottom right corner. Figure 9 for Figure 7 The brightness curve formed by pixels along the diagonal line from the bottom left corner to the top right corner. (Combined with...) Figure 7 , Figure 8 and Figure 9 As can be seen, after correction by the method of this invention, the brightness of the originally darker corner areas in the image is almost the same as that of the central area, indicating that this invention has a good correction effect on this type of microscopic image.

[0059] The present invention has been described in detail with reference to the foregoing embodiments. Those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for correcting vignetting in microscopic images, characterized in that, Includes the following steps: Step (1): Obtain the vignetting microscopic image to be processed, and convert the image from RGB color mode to HSV color mode microscopic image; decompose the HSV image into hue channel image H, saturation channel image S and luminance channel image V; Step (2): Construct and solve the optimization equation to obtain the brightness imbalance factor matrix L and the dark field layer D; The optimization equation is: in, Denotes the L1 norm of a matrix. G represents the corrected image. f (·) represents the Gaussian low-pass filter function, G m (·) denotes the median filter function, D m represents the average value of the dark field layer matrix D, min(V) represents the minimum value of the luminance channel image matrix V, and ⊙ represents element-wise multiplication of the matrices. This indicates the addition of corresponding elements of the matrix. Indicates matrix multiplication. To optimize the residual terms in the equation, and To optimize the regularization term in the equation, α1 and α2 are two regularization parameters; Step (3): Calculate the corrected image of the V channel image. Corrected image The original H hue channel image and S saturation channel image are recombined to form an HSV image; the resulting HSV image is then converted to an RGB image to complete the correction.

2. The method for correcting vignetting in microscopic images according to claim 1, characterized in that, In the optimization equation, the residual term constructs a function F(·) that is the Shannon-Wiener entropy, and the formula is: in, Represents the corrected image First use G m (·) Median filtering operation, followed by G(·) Gaussian low-pass filtering function operation. This represents the result after median and Gaussian filtering.

3. The method for correcting vignetting in microscopic images according to claim 2, characterized in that, The formula China G f (·) is the Gaussian low-pass filter function, G f The standard deviation σ in (·) is 3.2; the median filter function G m The template size for (·) is 3x3. express The value of the i-th pixel Probability, N is The number of pixels.

Images