A polarization image fusion method based on multi-scale transformation and image enhancement

Abstract

This invention discloses a polarization image fusion method based on multi-scale transformation and image enhancement. First, four polarization images with different polarization angles are acquired using a polarization camera, and the required polarization degree, polarization angle, and visible light image are calculated for synthesis. Then, the images are filtered and enhanced using fast guided filtering and enhanced CLAHE. Next, the polarization degree, polarization angle, and visible light image are decomposed into Laplacian image pyramids. For the high-frequency layers of the three Laplacian pyramid images, bilateral filtering and Laplacian sharpening are used to enhance high-frequency information. The brightness of the original image is calculated, and a brightness weight pyramid is generated. Weighted fusion is performed on each pyramid level. The fused Laplacian pyramid is reconstructed to obtain the final fused image. This invention can capture defect feature information from different dimensions of reflective metal surfaces and organically fuse them together, which has significant practical implications for subsequent defect detection, effectively improving the efficiency and accuracy of defect detection on reflective metal surfaces.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and more specifically, to a polarization image fusion method based on multi-scale transformation and image enhancement. Background Technology

[0002] Traditional methods for detecting defects on metal surfaces, such as visual inspection and ultrasonic testing, while effective in certain situations, often face challenges in detecting reflective metal surfaces. The strong reflection and scattering from reflective metal surfaces make it difficult to directly capture defect information, leading to decreased detection accuracy. Polarized light can be used to address the difficulty in capturing the characteristic features of defects on reflective metal surfaces.

[0003] Using polarization for image acquisition can suppress reflections to some extent. Furthermore, by acquiring polarized images at different angles, polarization degree and polarization angle images can be calculated. Compared to images acquired by ordinary cameras, these images contain information about metal surface defects in different dimensions. However, this information is scattered and may exist in different images; therefore, it needs to be integrated.

[0004] In the field of image processing, image fusion technology is widely used in various scenarios because it can integrate image information from different sensors or imaging methods. In the detection of defects on reflective metal surfaces, fusing image information from different scales and angles can significantly improve the accuracy and reliability of defect detection. A polarization degree image refers to the proportion of polarized light to the total light intensity, reflecting the degree of light polarization. Polarization degree images are less affected by uneven lighting or excessively low lighting, effectively improving the signal-to-noise ratio and contrast of images detected in complex lighting environments or under poor lighting conditions. A polarization angle image describes the polarization direction of light and reflects the surface reflection and scattering characteristics of the target object. Visible light images provide basic shape and color information of the target object, which is crucial for the initial identification and rapid localization of defects. Therefore, a polarization image fusion method based on multi-scale transformation and image enhancement is designed. Summary of the Invention

[0005] The technical problem to be solved by this invention is: to design a polarization image fusion method based on multi-scale transformation and image enhancement to solve the problem that defect information is difficult to capture and process due to light reflection and scattering when detecting defects on reflective metal surfaces.

[0006] To achieve the above objectives, the technical solution provided by this invention is: a polarization image fusion method based on multi-scale transformation and image enhancement, comprising the following steps:

[0007] Step 1: The polarization camera acquires polarization images at four different polarization angles, and the degree of polarization, polarization angle, and visible light image are obtained through calculation.

[0008] Step 2: Perform fast guided filtering and CLAHE enhancement on the polarization degree, polarization angle, and visible light images to improve image quality and contrast;

[0009] Step 3: Decompose the three types of images (degree of polarization, polarization angle, and visible light) into Laplacian image pyramids;

[0010] Step 4: Apply bilateral filtering to the high-frequency layer image in the Laplacian pyramid to smooth and preserve edge information, and apply Laplacian sharpening to enhance image details;

[0011] Step 5: Calculate the brightness of the original image and generate a brightness weight pyramid. Based on the brightness weight pyramid, perform weighted fusion on each pyramid level.

[0012] Step 6: Reconstruct the merged Laplace pyramid to obtain the final merged image.

[0013] Further, in step one, a polarization camera is used to acquire polarization images with polarization angles of 0°, 45°, 90°, and 135°, expressed using Stokes' theorem:

[0014]

[0015] In the formula, S0 is the total incident light intensity of the system, which is represented by the sum of the light intensities of linearly polarized light passing through the 0° and 90° directions; S1 is the difference in light intensities of linearly polarized light passing through the 0° and 90° directions; S2 is the difference in light intensities of linearly polarized light passing through the 45° and 135° directions; S3 is the difference in light intensities of right-handed and left-handed circularly polarized light; (x,y) represents the pixel coordinates.

[0016] Calculate the degree of polarization (Dop) image and the angle of polarization (Aop) image:

[0017]

[0018] Furthermore, in step two, fast guided filtering and CLAHE enhancement are applied to the polarization degree, polarization angle, and visible light image. The local linear filter for guided filtering is:

[0019]

[0020] Where q is the value of the output pixel, I is the value of the input image, i and k are pixel indices, and a and b are the coefficients of the linear function when the window center is at k.

[0021] Find the coefficients a of the linear function. k and b k The goal is to minimize the difference between the output value q of the fitted function and the true value p, which transforms the problem into an optimization problem, namely, minimizing the following expression:

[0022]

[0023] Using least squares, we can obtain:

[0024]

[0025] Where, μ k Is image I in window ω k The average value in The image to be filtered, p, is in window ω. k The mean of the middle, Is I in window ω k The variance in ω, |ω| is the window size ω k The quality of the mid-pixel, where ε is the regularization parameter.

[0026] To calculate the output value at a specific point, you need to average all the linear function values ​​that include that point, as follows:

[0027]

[0028] In the formula, ω k It represents all windows containing pixel i, and k is its center position.

[0029] Fast guided filtering, based on the guided filtering described above, downsamples the image to reduce the number of filter points, and then calculates the coefficient 'a'. k and b k After averaging the values, upsampling is performed to restore the original image size.

[0030] The core process of the enhanced CLAHE algorithm is as follows:

[0031] (1) Normally, the input image is divided into an 8*8 grid for calculation. Enhanced CLAHE first calculates the mean square error of the image brightness. The smaller the mean square error of the image brightness, the more consistent the brightness variation of the image is, and a larger number of grids, i.e., 8*8, is used; conversely, a smaller number of grids, i.e., 4*4, is used.

[0032] (2) Calculate the sub-block histogram.

[0033] (3) The histogram is clipped and redistributed, and then the clipped histogram is balanced to obtain the mapping table of each small block.

[0034] (4) Smoothing the mapping table. Perform one-dimensional Gaussian smoothing on the mapping table. For regions where the image transformation is relatively gentle, this smoothing can reduce the color blockiness caused by enhancement to a certain extent.

[0035] (5) Bilinear interpolation is performed on each small block to achieve the final enhancement effect.

[0036] Furthermore, in step three, the three types of images (degree of polarization, polarization angle, and visible light) are decomposed into Laplacian image pyramids. The specific process is as follows:

[0037] The polarization degree image, polarization angle image, and visible light image are used as the base G0 of the pyramid. Then, Gaussian low-pass filtering and downsampling are performed to construct a Gaussian pyramid of the image. The image G of the l-th layer of the Gaussian pyramid is... l for:

[0038]

[0039] In the formula, 1≤l≤N, 0<i<C l , 0 < j < R l N is the total number of layers, C l and R l For G l The number of rows and columns; ω(n,m) is the Gaussian kernel function.

[0040] Image G of each layer of the Gaussian pyramid l Image after subtracting upsampling and Gaussian convolution from the previous layer The resulting series of difference images constitutes the Laplacian pyramid (LP). l ,Right now:

[0041]

[0042] In the formula, ω(n,m) is the Gaussian kernel function, and G' is the kernel function when (i+m) and (j+n) are even. l+1 equals G l+1 In other cases, G' l+1 It is 0.

[0043] The Laplacian pyramid obtained by decomposing the polarization degree image, polarization angle image, and visible light image is denoted as Dop_LP. l Aop_LP l and I_LP l , where 0≤l≤N, and N is the number of decomposition layers.

[0044] Furthermore, in step four, bilateral filtering is applied to the high-frequency layer image in the Laplacian pyramid to smooth and preserve edge information, and Laplacian sharpening is applied to enhance image details.

[0045] The working principle of bilateral filtering is as follows:

[0046]

[0047] Where i and j are the coordinates of the current center pixel, g(i,j) is the filtered output, k and l are the coordinates of the neighboring pixels, f(k,l) is the neighboring pixel value, w(i,j,k,l) ​​is the weighting coefficient, and σd Let σ be the standard deviation of the Gaussian function of spatial location. r is the standard deviation of the Gaussian function for pixel similarity.

[0048] The Laplacian sharpening formula is:

[0049]

[0050] In the formula, g(i,j) is the sharpened image, f(i,j) is the unsharpened image, and the convolution kernel H(r,s) is...

[0051]

[0052] Further, in step five, the brightness of the original image is calculated to generate a brightness weight pyramid. Based on the brightness weight pyramid, weighted fusion is performed on each pyramid level.

[0053] Calculate the brightness of the image:

[0054]

[0055] Where E(i,j) is the calculated brightness image, I(i,j) is the value of the original image pixel after normalization to 0-1, and σ is the variance.

[0056] The polarization degree, polarization angle, and brightness image corresponding to the visible light image are decomposed into Gaussian pyramids, which are then used as fusion weights for the Laplacian pyramid of the original image. The calculated brightness weight pyramid is denoted as Dop_G. l Aop_G l and I_G l , where 0≤l≤N, and N is the number of decomposition layers.

[0057] The formula for calculating the fused image F(i,j) of each layer is:

[0058]

[0059] By merging each level of the pyramid in this way, a merged image of the Laplacian Pyramid (LP) is created. l F Where 0≤l≤N, and N is the number of decomposition layers.

[0060] Furthermore, in step six, the merged Laplace pyramid is reconstructed to obtain the final merged image.

[0061] After fusing the three images using the method described above, a Laplacian pyramid of the fused image is formed. Reconstruction of the Pyramid of Gauss from the Pyramid of Laplace The resulting underlying image This is the final fused image, expressed as:

[0062]

[0063] In the formula, ω(n,m) is the Gaussian kernel function. Given an upsampled Gaussian pyramid image, when (i+m) and (j+n) are even numbers... equal In other cases It is 0.

[0064] Compared with existing methods, the advantages of this invention are as follows: It proposes a polarization image fusion method based on multi-scale transformation and image enhancement. Compared with traditional methods, it achieves the capture of defect feature information of reflective metals from different dimensions, overcoming the difficulty of obtaining clear images under complex lighting conditions. Before and during the multi-scale transformation, image enhancement is performed multiple times to enhance edge and detail information in the image, especially the defect parts, making the shape and location of defects clearer and significantly improving the accuracy of defect detection. The brightness-weighted fusion strategy adopted in the fusion stage dynamically adjusts the fusion weights of each image layer according to the fusion source image, ensuring that the fusion process highlights key high-brightness defect objects while preserving background and detail information in low-brightness areas. This has significant practical implications for subsequent defect detection, effectively improving the efficiency and accuracy of defect detection on reflective metal surfaces. Attached Figure Description

[0065] Figure 1 This is the overall flowchart of the present invention;

[0066] Figure 2 (a), (b), and (c) are the polarization degree, polarization angle, and visible light images calculated by the present invention, respectively.

[0067] Figure 3 (a), (b), and (c) are the polarization degree, polarization angle, and visible light images after the fast guided filtering of this invention;

[0068] Figure 4 (a), (b), and (c) are the enhanced polarization degree, polarization angle, and visible light images of the present invention, respectively.

[0069] Figure 5 (a), (b), and (c) are respectively the brightness images corresponding to the degree of polarization, polarization angle, and visible light image generated by the present invention;

[0070] Figure 6 This is the fused image obtained by the present invention. Detailed Implementation

[0071] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the embodiments of the present invention.

[0072] Example:

[0073] See Figure 1 The polarization image fusion method based on multi-scale transformation and image enhancement in this embodiment includes the following steps:

[0074] Step 1: Based on the polarization images acquired by the polarization camera at four different polarization angles, synthesize the polarization degree, polarization angle, and visible light image. The polarization camera acquired polarization images with polarization angles of 0°, 45°, 90°, and 135°, and the relevant parameters were calculated using Stokes' theorem.

[0075]

[0076] In the formula, S0 is the total incident light intensity of the system, which is represented by the sum of the light intensities of linearly polarized light passing through the 0° and 90° directions; S1 is the difference in light intensities of linearly polarized light passing through the 0° and 90° directions; S2 is the difference in light intensities of linearly polarized light passing through the 45° and 135° directions; S3 is the difference in light intensities of right-handed and left-handed circularly polarized light; (x,y) represents the pixel coordinates.

[0077] Calculate the degree of polarization (Dop) image and the angle of polarization (Aop) image:

[0078]

[0079] The polarization degree, polarization angle, and visible light image of this embodiment are referenced. Figure 2 .

[0080] Step two, use fast guided filtering and enhanced CLAHE to filter and enhance the polarization degree, polarization angle, and visible light image. The local linear filter for guided filtering is:

[0081]

[0082] Where q is the value of the output pixel, I is the value of the input image, i and k are pixel indices, and a and b are the coefficients of the linear function when the window center is at k.

[0083] Find the coefficients a of the linear function. k and b k The goal is to minimize the difference between the output value q of the fitted function and the true value p, which transforms the problem into an optimization problem, namely, minimizing the following expression:

[0084]

[0085] Using least squares, we can obtain:

[0086]

[0087]

[0088] Where, μ k Is image I in window ω k The average value in The image to be filtered, p, is in window ω. k The mean of the middle, Is I in window ω k The variance in ω, |ω| is the window size ω k The quality of the mid-pixel, where ε is the regularization parameter.

[0089] To calculate the output value at a specific point, you need to average all the linear function values ​​that include that point, as follows:

[0090]

[0091] In the formula, ω k It represents all windows containing pixel i, and k is its center position.

[0092] Fast guided filtering, based on the guided filtering described above, downsamples the image to reduce the number of filter points, and then calculates the coefficient 'a'. k and b k After averaging the values, upsampling is performed to restore the original image size. In this embodiment, the downsampling factor is 4, and the window size is 5×5.

[0093] In this embodiment, the polarization degree, polarization angle, and visible light image after fast guided filtering are shown in the reference document. Figure 3 .

[0094] The core process of the enhanced CLAHE algorithm is as follows:

[0095] (1) Normally, the input image is divided into an 8*8 grid for calculation. Enhanced CLAHE first calculates the mean square error of the image brightness. The smaller the mean square error of the image brightness, the more consistent the brightness variation of the image is, and a larger number of grids, i.e., 8*8, is used; conversely, a smaller number of grids, i.e., 4*4, is used.

[0096] (2) Calculate the sub-block histogram.

[0097] (3) The histogram is clipped and redistributed, and then the clipped histogram is balanced to obtain the mapping table of each small block.

[0098] (4) Smoothing the mapping table. Perform one-dimensional Gaussian smoothing on the mapping table. For regions where the image transformation is relatively gentle, this smoothing can reduce the color blockiness caused by enhancement to a certain extent.

[0099] (5) Bilinear interpolation is performed on each small block to achieve the final enhancement effect.

[0100] For the enhanced polarization degree, polarization angle, and visible light image of this embodiment, please refer to [link / reference]. Figure 4 .

[0101] Step 3: Decompose the polarization degree, polarization angle, and visible light image into a Laplacian image pyramid. Use the three original images as the base G0 of the pyramid, and then perform Gaussian low-pass filtering and downsampling to construct the Gaussian pyramid of the image. The first layer of the Gaussian pyramid is image G0. l for:

[0102]

[0103] In the formula, 1≤l≤N, 0<i<C l , 0 < j < R l N is the total number of layers, C l and R l For G l The number of rows and columns; ω(n,m) is the Gaussian kernel function. In this embodiment, log(min(rows,cols)) / log(2) is taken, where rows is the number of rows of image pixels and cols is the number of columns of image pixels. The number of rows is rows = 1024 and the number of columns is cols = 1224, so N = 10 is calculated.

[0104] Image G of each layer of the Gaussian pyramid l Image after subtracting upsampling and Gaussian convolution from the previous layer The resulting series of difference images constitutes the Laplacian pyramid (LP). l ,Right now:

[0105]

[0106] In the formula, ω(n,m) is the Gaussian kernel function, and G' is the kernel function when (i+m) and (j+n) are even. l+1 equals G l+1 In other cases, G' l+1 It is 0.

[0107] The Laplacian pyramid obtained by decomposing the polarization degree image, polarization angle image, and visible light image is denoted as Dop_LP. l Aop_LP l and I_LP l , where 0≤l≤N, N is the number of decomposition layers. In this embodiment, log(min(rows,cols)) / log(2) is taken, where rows is the number of rows of image pixels and cols is the number of columns of image pixels. The number of rows is rows=1024 and the number of columns is cols=1224. Therefore, N=10 is calculated.

[0108] Step four: Perform bilateral filtering on the high-frequency layer image in the Laplacian pyramid to smooth and preserve edge information, and apply Laplacian sharpening to enhance image details.

[0109] The formula for bilateral filtering is as follows:

[0110]

[0111] Where i and j are the coordinates of the current center pixel, g(i,j) is the filtered output, k and l are the coordinates of the neighboring pixels, f(k,l) is the neighboring pixel value, w(i,j,k,l) ​​is the weighting coefficient, and σ d Let σ be the standard deviation of the Gaussian function of spatial location. r is the standard deviation of the Gaussian function for pixel similarity.

[0112] The Laplacian sharpening formula is:

[0113]

[0114] In the formula, g(i,j) is the sharpened image, f(i,j) is the unsharpened image, and the convolution kernel H(r,s) is...

[0115]

[0116] Step 5: Calculate the brightness of the original image and generate a brightness weight pyramid. Based on the brightness weight pyramid, perform weighted fusion on each pyramid level.

[0117] Calculate the brightness of the image:

[0118]

[0119] Where E(i,j) is the calculated brightness image, I(i,j) is the value of the original image pixel after normalization to 0-1, and σ is the variance, which is 0.2 in this embodiment.

[0120] The polarization degree, polarization angle, and brightness image corresponding to the visible light image generated in this embodiment are shown in the reference document. Figure 5 .

[0121] The polarization degree, polarization angle, and brightness image corresponding to the visible light image are decomposed into Gaussian pyramids, which are then used as fusion weights for the Laplacian pyramid of the original image. The calculated brightness weight pyramid is denoted as Dop_G. l Aop_G l and I_G l , where 0≤l≤N, and N is the number of decomposition layers.

[0122] The formula for calculating the fused image F(i,j) of each layer is:

[0123]

[0124] By merging each level of the pyramid in this way, a merged image of the Laplacian Pyramid (LP) is created. l F Where 0≤l≤N, and N is the number of decomposition layers. In this embodiment, log(min(rows,cols)) / log(2) is used, where rows is the number of rows of image pixels and cols is the number of columns of image pixels. The number of rows is rows=1024 and the number of columns is cols=1224, so N=10 is calculated.

[0125] Step six: Reconstruct the merged Laplace pyramid to obtain the final merged image.

[0126] After fusing the three images using the method described above, a Laplacian pyramid of the fused image is formed. Reconstruction of the Pyramid of Gauss from the Pyramid of Laplace The resulting underlying image This is the final fused image, expressed as:

[0127]

[0128]

[0129] In the formula, ω(n,m) is the Gaussian kernel function. Given an upsampled image of the Gaussian pyramid, when (i+m) and (j+n) are even numbers, equal In other cases It is 0.

[0130] In this embodiment, the fused image is referred to Figure 6 .

[0131] The above description of the present invention and its embodiments is not restrictive. The accompanying drawings are merely one embodiment of the present invention, and the actual application is not limited thereto. If those skilled in the art, inspired by this description, design similar structures and embodiments without departing from the spirit of the present invention, such designs should fall within the protection scope of the present invention.

Claims

1. A polarization image fusion method based on multi-scale transformation and image enhancement, characterized in that, The method includes the following steps: Step 1: The polarization camera acquires polarization images at four different polarization angles, and the degree of polarization, polarization angle, and visible light image are obtained through calculation. Step 2: Perform fast guided filtering and enhanced CLAHE enhancement on the polarization degree, polarization angle, and visible light images to improve image quality and contrast; The core process of the enhanced CLAHE algorithm is as follows: (1) Enhanced CLAHE first calculates the mean square error of image brightness. The smaller the mean square error of image brightness, the more consistent the brightness variation of the image, and the more grid number is used, i.e., 8*8; conversely, the fewer grid number is used, i.e., 4*4. (2) Calculate the sub-block histogram. (3) The histogram is clipped and redistributed, and then the clipped histogram is balanced to obtain a mapping table for each small block. (4) Smoothing the mapping table: Gaussian smoothing is applied to the mapping table in one dimension. For areas where the image transformation is relatively gentle, this smoothing can reduce the color blockiness caused by enhancement to a certain extent. (5) Perform bilinear interpolation on each small block to achieve the final enhancement effect; Step 3: Decompose the three images—degree of polarization, polarization angle, and visible light—into a Laplacian image pyramid; Step 4: Apply bilateral filtering to the high-frequency layer image in the Laplacian pyramid to smooth and preserve edge information, and apply Laplacian sharpening to enhance image details; Step 5: Calculate the brightness of the original image, generate a brightness weight pyramid, and perform weighted fusion on each pyramid level based on the brightness weight pyramid; Step 6: Reconstruct the merged Laplace pyramid to obtain the final merged image.

2. The polarization image fusion method based on multi-scale transformation and image enhancement according to claim 1, characterized in that, In step one, the polarization camera acquires polarization images at four different polarization angles. The degree of polarization, polarization angle, and visible light image are then calculated. The specific operation is as follows: Polarized images with polarization angles of 0°, 45°, 90°, and 135° were acquired using a polarization camera, expressed using Stokes' theorem: In the formula, The total incident light intensity of the system is represented by the sum of the intensities of linearly polarized light passing through the 0° and 90° directions; The difference in intensity of linearly polarized light at 0° and 90°. The difference between linearly polarized light at 45° and 135°. This represents the difference in intensity between right-handed and left-handed circularly polarized light. Represents pixel coordinates. Calculate the degree of polarization image Dop and the angle of polarization image Aop: 。 3. The polarization image fusion method based on multi-scale transformation and image enhancement according to claim 1, characterized in that, The specific operations for preprocessing the degree of polarization, polarization angle, and visible light image in step two are as follows: The local linear filter for guided filtering is: in, It outputs the pixel value. It is the value of the input image. and It is a pixel index. and When the center of the window is located The coefficients of the linear function at that time. Find the coefficients of the linear function and That is, to make the output value of the fitting function and the true value To minimize the difference between them, we transform the problem into an optimization problem, which means minimizing the following expression: Using least squares, we get: in, It is an image In the window The average value in The image to be filtered In the window The mean of the middle, yes In the window The variance in It is a window Medium pixel quality, For regularization parameters, To calculate the output value at a specific point, you need to average all the linear function values ​​that include that point, as follows: In the formula, It is all pixels The window, It is its central location. Fast guided filtering, based on the guided filtering described above, downsamples the image to reduce the number of filter points, and then calculates the coefficients. and After averaging the values, upsampling is performed to restore the original image size.

4. The polarization image fusion method based on multi-scale transformation and image enhancement according to claim 1, characterized in that, Step three involves decomposing the polarization degree, polarization angle, and visible light images into Laplacian image pyramids. The specific steps are as follows: Using polarization degree image, polarization angle image, and visible light image as the base of the tower Then, Gaussian low-pass filtering and downsampling are performed to construct the Gaussian pyramid of the image. Layer Image for: In the formula, , , ; This represents the total number of floors. and for The number of rows and columns; For Gaussian kernel function, Images of each level of the Gaussian pyramid Image after subtracting upsampling and Gaussian convolution from the previous layer The resulting series of difference images constitutes the Laplace pyramid. ,Right now: In the formula, Let be the Gaussian kernel function, when and When it is even, equal In other cases, =0, The Laplacian pyramid obtained by decomposing the polarization degree image, polarization angle image, and visible light image is denoted as: , and ,in , The number of layers in the decomposition.

5. The polarization image fusion method based on multi-scale transformation and image enhancement according to claim 1, characterized in that, In step four, bilateral filtering is applied to the high-frequency layer image in the Laplacian pyramid to smooth and preserve edge information, and Laplacian sharpening is applied to enhance image details. The specific operations are as follows: The working principle of bilateral filtering is as follows: in and These are the coordinates of the current center pixel. This is the filtered output. and These are the coordinates of neighboring pixels. The value of the neighboring pixels. These are weighting coefficients. The standard deviation of the Gaussian function of spatial location. The standard deviation of the Gaussian function for pixel similarity. The Laplacian sharpening formula is: In the formula, The image after sharpening. For the image before sharpening, the convolution kernel... for 。 6. The polarization image fusion method based on multi-scale transformation and image enhancement according to claim 1, characterized in that, Step five involves calculating the brightness of the original image and generating a brightness weight pyramid. The specific operation of weighted fusion at each pyramid level based on the brightness weight pyramid is as follows: Calculate the brightness of the image: in, This is the brightness image after calculation. These are the values ​​of the original image pixels after normalization to 0-1. It is variance. The polarization degree, polarization angle, and brightness image corresponding to the visible light image are decomposed into Gaussian pyramids, which are used as the fusion weights of the original image's Laplacian pyramid. The calculated brightness weight pyramid is denoted as... , and ,in , The number of layers in the decomposition. fused image of each layer The calculation formula is: By merging each level of the pyramid in this way, a fused image of the Pyramid of Laplace is created. ,in , The number of layers in the decomposition.

7. The polarization image fusion method based on multi-scale transformation and image enhancement according to claim 1, characterized in that, Step six involves reconstructing the merged Laplacian pyramid to obtain the final merged image. The specific steps are as follows: After fusing the three images using the method described above, a Laplacian pyramid of the fused image is formed. , … The Pyramid of Gauss was reconstructed from the Pyramid of Laplace. , … The resulting underlying image This is the final fused image, expressed as: In the formula, For Gaussian kernel function, For an upsampled image of a Gaussian pyramid, when and When it is even, equal In other cases, It is 0.

Images