Multi-scale image enhancement fusion method and system
Through the multi-scale image enhancement fusion method, combining intensity and polarization information, using frequency decomposition and adaptive weight adjustment, the problems of poor image quality and poor depth of field recovery in the prior art are solved, and more comprehensive and flexible image processing is achieved , ensuring the overall quality and detail retention of the image.
Patent Information
- Application Number
- CN202410994536.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-24
- Publication Date
- 2024-11-15
AI Technical Summary
The prior art has limitations in image enhancement, especially in haze environments, which makes it difficult to effectively utilize polarization information, resulting in poor image quality and poor depth of field recovery effect. The traditional method ignores the fusion of different frequency information, resulting in images Partially over-or inadequate enhancement.
Using multi-scale image enhancement fusion method, through multi-scale enhancement fusion of intensity and polarization information, low-frequency and high-frequency information are obtained by frequency decomposition of two-dimensional image features, and edge-retaining characteristics and adaptive weight adjustment strategies are used to fusion images to ensure that Balance processing of information within different frequency ranges of the image.
Achieve image quality improvement, blending more edge and detail features, ensuring proper enhancement of target features and details, while maintaining depth of field information without causing excessive or insufficient enhancement, improving the overall quality and robustness of the image sex.
Smart Images

Figure CN118967470A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of image processing technology, and in particular to a multi-scale image enhancement fusion method and system. Background Art
[0002] In scene-based optical detection, optoelectronic imaging is an important technology used to obtain real original images. However, in non-ideal weather conditions, these images are often interfered by low visibility and stray light, which reduces the visibility of scene targets. In particular, in haze environments, scattered light from particles and water droplets will produce background stray light, which makes scene targets more difficult to distinguish, resulting in poor image quality and limiting its application in scene optical detection. In addition, low visibility may cause distant targets to be blurred, further reducing the amount of information in the image.
[0003] Polarization is a basic property of light, which describes the relationship between the propagation direction and vibration direction of light waves, and contains special physical intrinsic information of the target. Polarization imaging is a technology that uses the polarization properties of light to acquire and analyze images. Unlike traditional intensity images, polarization imaging can obtain the polarization state information of light reflected or radiated from the surface of an object, and obtain additional information about the shape, direction, material properties, etc. of the target, thereby providing richer visual features and surface information, enhancing the difference between the target and the natural background, and is of great significance for target detection. Compared with light intensity images, polarization imaging in complex scenes has the advantages of improving target contrast and details, ambient light robustness, and object surface characteristics analysis. In haze weather, images may be scattered and absorbed by haze particles, resulting in image blur and reduced contrast. Polarization imaging analyzes the vibration direction and intensity of light under different polarization directions, obtains information about the material properties, texture and shape of the object surface, reduces the interference of haze particles, and provides clearer and more accurate image information.
[0004] Existing technologies have limitations in image enhancement. Existing image enhancement methods usually focus on the processing of intensity images, while ignoring the potential value of polarization information. Although intensity images provide information on the light intensity distribution on the surface of an object, under certain meteorological conditions, such as haze, they are limited by visibility and it is difficult to provide a clear target outline. In this case, polarization information can provide more information about specific features of the target, such as reflectivity, material, and shape. Therefore, the failure to fully utilize polarization data limits the effectiveness of image enhancement. Existing image enhancement methods usually ignore the fusion of information at different frequencies. Images contain multiple frequency components, from low-frequency global structural information to high-frequency detail information. Traditional enhancement methods have difficulty in effectively distinguishing and processing these different frequency information, resulting in over-enhancement or under-enhancement of some parts of the image. This unbalanced processing method limits the overall quality of the image. Summary of the invention
[0005] In order to overcome the shortcomings of the prior art, the present invention provides a multi-scale image enhancement fusion method and system to solve the problems of poor image enhancement fusion quality and poor depth of field restoration effect existing in the prior art.
[0006] The technical solution adopted by the present invention to solve the above problems is:
[0007] A multi-scale image enhancement fusion method is used to enhance the image by multi-scale enhancement fusion of intensity and polarization information.
[0008] As a preferred technical solution, the following steps are included:
[0009] S1, data acquisition: build a polarization imaging detection system to simultaneously collect intensity images of multiple polarization angles in real scenes;
[0010] S2, representation and solution: Based on the intensity image representation and solution, a multi-parameter polarization image is obtained, the polarization image and the intensity image are selected for image alignment, and the aligned image is input into the enhanced fusion system as an input image;
[0011] S3, decomposition: Use the frequency of the two-dimensional image features to decompose the intensity image and the polarization image respectively to obtain information in different frequency ranges, and the information in different frequency ranges includes low-frequency sub-bands and high-frequency sub-bands;
[0012] S4, low-frequency subband creation: using the edge preservation feature, the weighted average of the pixel values in the low-frequency subband is calculated, the retained pixels are selected, and a new low-frequency subband is created;
[0013] S5, image fusion: the intensity image is fused with the low-frequency subband of the polarization image. The fusion weight is determined by the image gradient value, energy map information and empirical weight adjustment factor. The adaptive weight adjustment is completed by constructing the weight equation.
[0014] S6, maximum fusion: perform maximum fusion on the high-frequency sub-bands and select the larger value of the pixels at the corresponding positions of the two images;
[0015] S7, reconstruction: reconstruct the low-frequency sub-band and the high-frequency sub-band using the inverse transform of the image features to obtain a fused image.
[0016] As a preferred technical solution, in step S3, the intensity image and the polarization image are decomposed by discrete wavelet transform to obtain information in different frequency ranges;
[0017] The information of different frequency ranges includes low frequency sub-band cA and high frequency sub-band. The high frequency sub-band includes horizontal high frequency sub-band cH, vertical high frequency sub-band cV and diagonal high frequency sub-band cD:
[0018]
[0019] Among them, cA j [k] represents the j-th level approximation subband, cD j [k] represents the j-th level detail subband, k represents the index of the subband, j represents the level of transformation, n represents the index variable in the transformation process, N represents the total length, ψ(·) represents the wavelet basis function, φ(·) represents the scaling function, and · represents the convolution operation.
[0020] As a preferred technical solution, in step S4, the formula for selecting the retained pixels is:
[0021]
[0022]
[0023] Where I(q) represents the input image, G(p) represents the output image, q represents the image pixel coordinates of the input image, p represents the image pixel coordinates of the output image, σ s represents the standard deviation, σ r represents the spatial standard deviation of the control space range, N(p) represents the filter window centered at pixel p, and K p represents the normalization factor, J t (p) represents the result after the t-1th iteration, J t+1 (p) represents the result after the tth iteration.
[0024] As a preferred technical solution, step S5 includes the following steps:
[0025] S51, high frequency sub-band fusion: perform maximum value fusion on the horizontal high frequency sub-band cH, the vertical high frequency sub-band cV and the diagonal high frequency sub-band cD, and select the larger value of the pixels at the corresponding positions of the two images as the fusion result.
[0026] As a preferred technical solution, step S5 includes the following steps:
[0027] S52, low frequency sub-band fusion: the low frequency sub-bands are fused by weighted averaging.
[0028] As a preferred technical solution, in step S52, when the low-frequency sub-bands are fused, the calculation formula of the fusion weight is:
[0029] α=Sig[ω]=Sig[(1-β)×G(i,j)+β×E(i,j)]
[0030] Among them, α represents the fusion weight, ω represents the weight function, Sig[·] represents the mapping relationship, β represents the weight adjustment factor, (i, j) represents the pixel coordinates, G(i, j) represents the gradient map information of the image, and E(i, j)] represents the energy map information of the image.
[0031] As a preferred technical solution, in step S2, the intensity image is calculated separately to obtain a multi-parameter polarization image after the original polarization image is solved, and the parametric polarization image and the intensity image are selected together as the input image for image enhancement fusion.
[0032] As a preferred technical solution, the method further comprises the following steps:
[0033] S7, equalization: histogram equalization: perform histogram equalization on the intensity image, polarization image, and fused image.
[0034] An enhanced fusion system, used to implement the multi-scale image enhanced fusion method, comprises the following modules connected in sequence:
[0035] Data acquisition module: used to build a polarization imaging detection system and simultaneously collect intensity images of multiple polarization angles in real scenes;
[0036] Characterization and solution module: used to obtain a multi-parameter polarization image based on the intensity image characterization solution, select the polarization image and the intensity image for image alignment, and input the aligned image as an input image into the enhanced fusion system;
[0037] Decomposition module: used to decompose the intensity image and the polarization image respectively using the frequency of the two-dimensional image features to obtain information in different frequency ranges, where the information in different frequency ranges includes low-frequency sub-bands and high-frequency sub-bands;
[0038] Low-frequency sub-band creation module: used to calculate the weighted average of pixel values in the low-frequency sub-band by utilizing the edge preservation characteristic, select the retained pixels, and create a new low-frequency sub-band;
[0039] Image fusion module: used to fuse the intensity image with the low-frequency subband of the polarization image. The fusion weight is determined by the image gradient value, energy map information and empirical weight adjustment factor. The adaptive weight adjustment is completed by constructing the weight equation.
[0040] Maximum fusion module: used to perform maximum fusion on high-frequency sub-bands and select the larger value of pixels at corresponding positions of the two images;
[0041] Reconstruction module: used to reconstruct the low-frequency sub-band and the high-frequency sub-band using the inverse transformation of image features to obtain a fused image.
[0042] Compared with the prior art, the present invention has the following beneficial effects:
[0043] (1) The present invention integrates multi-source characterization information of intensity and polarization degree at different scales, and integrates more edge and detail features to enhance image quality;
[0044] (2) The present invention adopts an adaptive fusion weight adjustment strategy to make image enhancement more flexible and adaptable to different scenarios;
[0045] (3) The present invention adopts a multi-scale image enhancement fusion scheme. This strategy allows the image to be decomposed and enhanced in different frequency ranges, and the enhanced image fusion is performed to generate the best matching result. In this way, the important information of the image can be processed more comprehensively to ensure that the specific features and details of the target are properly enhanced without losing or damaging important depth of field information. BRIEF DESCRIPTION OF THE DRAWINGS
[0046] Figure 1 It is a flow chart of an embodiment of a physically intrinsic correlation multi-scale enhancement fusion method applicable to polarization degree images of the present invention;
[0047] Figure 2 Schematic diagram for obtaining fusion weights. DETAILED DESCRIPTION
[0048] The present invention will be further described in detail below in conjunction with embodiments and drawings, but the embodiments of the present invention are not limited thereto.
[0049] Example 1
[0050] like Figure 1 to Figure 2 As shown, the present invention relates to the field of polarization imaging detection and image processing, and in particular to an enhanced fusion technology of polarization images in complex scenes, and belongs to the application of polarization imaging in the field of imaging in scattering media.
[0051] In order to overcome the shortcomings of the prior art, the present invention discloses a physically intrinsically correlated multi-scale enhancement fusion method suitable for polarization images. By fusing intensity and polarization information, we can improve the quality of the image, especially under non-ideal weather and complex background conditions such as low visibility and stray light. This helps to improve the detection efficiency of targets in complex scenes. In addition, the use of a multi-scale image enhancement fusion method can better balance the information in different frequency ranges in the image. This means that both global structures and local details can be appropriately enhanced without causing problems of over- or under-enhancement. By introducing an adaptive weight strategy, the fusion weights are automatically adjusted according to the gradient and energy of the image, greatly enhancing the robustness and adaptability of the fusion.
[0052] This patent discloses an innovative method for enhancing image quality using polarization information. The method processes raw image data through a series of steps. First, the raw image includes intensity images of four polarization angles and the polarization degree image is solved. Subsequently, the images are decomposed using the frequency of image features to obtain subbands containing different frequency information. In the low-frequency subband, the pixel values are weighted averaged by using multi-source polarization degree images and retaining edge feature information to generate new low-frequency subbands. The low-frequency subbands are fused by weighted average, and the fusion weights are controlled by an adaptive weight adjustment strategy. The high-frequency subbands are fused by the maximum value method. Finally, the fused subbands are reconstructed by inverse transformation of image features to obtain the final fused image. After the image fusion is completed, both the intensity image and the polarization degree image are histogram equalized to further improve the image quality. This patent application describes a physically intrinsic correlation multi-scale enhancement fusion method suitable for polarization degree images, which can improve the contrast between the target and the background in complex scenes. Its adaptive fusion and information enhancement characteristics are suitable for a variety of applications. This innovative method performs well in processing detection images in foggy road environments and provides a powerful tool for improving visualization effects.
[0053] The present invention studies a new algorithm for inverting intensity information and polarization information by utilizing the differential evolution law of polarization images, thereby realizing the separation of target light and scattered light; the present invention studies the combination of physical intrinsic models and digital image enhancement, and performs high-frequency extraction in the frequency domain according to the regular structural characteristics of the target in the scene polarization image, thereby further improving the contrast between the target and the background, and ultimately achieving a significant improvement in the target detection effect.
[0054] The present invention is implemented by the following technical solutions:
[0055] A physically intrinsic correlation multi-scale enhancement fusion method applicable to polarization degree images, the method comprising the following steps:
[0056] Step 1: Build a polarization imaging detection system to collect intensity images at four polarization angles.
[0057] Step 2: For the collected original intensity image, the intensity image at each angle (0 degree, 45 degree, 90 degree, 135 degree) is calculated separately to obtain the multi-parameter polarization image after the original polarization image is solved, the multi-parameter polarization image light intensity image S0, the linear polarization component S1 in the horizontal or vertical direction, the linear polarization component S2 in the direction of +45° or -45°, the polarization angle image AoP and the polarization image DoLP, etc. Select the appropriate parametric polarization image and intensity image to complete the image alignment, and input them into the enhanced fusion system as the input image.
[0058] Step 3, using the frequency of the two-dimensional image features, respectively decompose the intensity image and the polarization image to obtain information in different frequency ranges, including low-frequency sub-bands and high-frequency sub-bands.
[0059] Step 4, using the edge preservation feature, calculate the weighted average of the pixel values in the low-frequency sub-band, select the retained pixels, improve the edge information of the target information, and create a new low-frequency sub-band.
[0060] Step 5: The intensity image is fused with the low-frequency sub-band of the polarization image. The fusion weight is used to construct a weight equation based on the image gradient and energy information and the empirical weight influencing factor to complete the adaptive weight adjustment.
[0061] Step 6: Perform maximum value fusion on the high-frequency sub-bands, that is, select the larger value of the pixels at corresponding positions of the two images.
[0062] Step 7: Reconstruct the sub-bands using inverse transform of image features to obtain the final fused image.
[0063] Step 8: Perform histogram equalization on the original intensity image, polarization degree image and fused image to improve image quality.
[0064] Compared with the prior art, the present invention discloses a physically intrinsic correlation multi-scale enhancement fusion method applicable to polarization degree images, which can achieve the following advantages:
[0065] (1) The present invention integrates multi-source characterization information of intensity and polarization degree, and integrates more edge and detail features to enhance image quality;
[0066] (2) Adopting an adaptive fusion weight adjustment strategy to make image enhancement more flexible and adaptable to different scenarios;
[0067] (3) A multi-scale image enhancement fusion scheme is adopted. This strategy allows the image to be decomposed and enhanced in different frequency ranges, and the enhanced images are fused to generate the best matching result. In this way, the important information of the image can be processed more comprehensively, ensuring that the specific features and details of the target are appropriately enhanced without losing or damaging important depth of field information.
[0068] Example 2
[0069] like Figure 1 to Figure 2 As shown, as a further optimization of Example 1, based on Example 1, this embodiment also includes the following technical features:
[0070] A multi-scale enhancement fusion method for physically intrinsic correlation of polarization images in foggy weather, comprising the following steps:
[0071] Step 1: Data collection: In haze weather conditions, use polarization imaging equipment to collect real scene original images. This process includes obtaining intensity images at the required four analysis angles.
[0072] Step 2: Polarization degree image calculation: For the collected original intensity image, the intensity image at each angle (0 degree, 45 degree, 90 degree, 135 degree) is calculated separately to obtain the multi-parameter polarization degree image after the original polarization degree image is calculated, including the multi-parameter polarization degree image light intensity image S0, the linear polarization component S1 in the horizontal or vertical direction, the linear polarization component S2 in the direction of +45° or -45°, the polarization angle image AoP and the polarization degree image DoLP, etc. These multi-polarization degree images will be used together with the intensity image as the input image for image enhancement fusion.
[0073] Step 3: Frequency decomposition of image features: Through discrete wavelet transform decomposition, the intensity image and polarization image are decomposed to obtain information in different frequency ranges, including low-frequency sub-bands cA and high-frequency sub-bands, namely horizontal cH, vertical cV and diagonal cD sub-bands.
[0074]
[0075] Among them, cA j [k] and cD j [k] are the j-th approximation coefficients and detail coefficients, respectively. k is the index of the subband and j is the number of transformations. ψ and φ are the wavelet basis function and the scaling function, and (·) represents the convolution operation.
[0076] Step 4: Rolling-guided edge preservation: Using the rolling-guided edge preservation feature, traverse each window of the low-frequency subband, calculate the weighted average of the pixel values in the window, select the pixel with the largest weighted average result as the retained pixel, and place it in the new low-frequency subband (cAnew).
[0077]
[0078] Where I(q) represents the input image, G(p) represents the output image, p and q represent the corresponding image pixel coordinates, σ s is the standard deviation, and N(p) represents the filter window centered at pixel p. t+1 (p) is the result of the tth iteration. In the iterative operation process, the J generated in the previous iteration of I(q) is given. t (p), we can get the result value J after the tth iteration t+1 (p).
[0079] Step 5: Multi-scale fusion: In this step, multi-scale fusion is performed to fuse the low-frequency sub-bands and the corresponding high-frequency sub-bands. This includes the following two sub-steps:
[0080] Step 5.1: High-frequency sub-band fusion: Perform maximum value fusion on the high-frequency sub-bands (cH, cV, cD), that is, select the larger value of the pixels at the corresponding positions of the two images as the fusion result.
[0081] Step 5.2: Low-frequency subband fusion: The low-frequency subbands are fused by weighted averaging, where the fusion weight α is controlled by a parameter. The fusion weight α adopts an adaptive weight adjustment strategy based on image gradient and energy to achieve better fusion results. β is an empirical weight adjustment factor, which is obtained mainly by mathematically counting the images in the acquired data set through formula (4) and is taken as 0.425.
[0082] α=Sig[ω]=Sig[(1-β)×G(i,j)+β×E(i,j)](4)
[0083] Step 6: Reconstruction: Use the inverse transform of image features to reconstruct the fused sub-bands (cAnew, cH, cV, cD) to obtain the final fused image (img_fused).
[0084] Step 7: Histogram equalization: Perform histogram equalization on the intensity image (img1), polarization image (img2) and fused image to improve image quality and contrast.
[0085] The advantage of this implementation is that it combines the intensity and polarization information, uses multi-scale image enhancement fusion technology to improve image quality, and decomposes the edge and detail features in the image for precise enhancement. At the same time, it adopts an adaptive fusion weight strategy, which is applicable to a wider range of scene image processing, improves the generalization ability of the fusion scheme, and thus provides a more effective scene image enhancement method.
[0086] As described above, the present invention can be preferably implemented.
[0087] All features disclosed in all embodiments in this specification, or steps in all methods or processes implicitly disclosed, except for mutually exclusive features and / or steps, can be combined and / or expanded or replaced in any manner.
[0088] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. According to the technical essence of the present invention, within the spirit and principles of the present invention, any simple modification, equivalent replacement and improvement made to the above embodiment still falls within the protection scope of the technical solution of the present invention.
Claims
1. A multi-scale image enhancement fusion method, characterized in that: Image enhancement is performed through multi-scale enhancement fusion of intensity and polarization information.
2. The multi-scale image enhancement fusion method according to claim 1, characterized in that: The following steps are involved: S1, data acquisition: build a polarization imaging detection system to simultaneously collect intensity images of multiple polarization angles in real scenes; S2, representation and solution: Based on the intensity image representation and solution, a multi-parameter polarization image is obtained, the polarization image and the intensity image are selected for image alignment, and the aligned image is input into the enhanced fusion system as an input image; S3, decomposition: Use the frequency of the two-dimensional image features to decompose the intensity image and the polarization image respectively to obtain information in different frequency ranges, and the information in different frequency ranges includes low-frequency sub-bands and high-frequency sub-bands; S4, low-frequency subband creation: using the edge preservation feature, the weighted average of the pixel values in the low-frequency subband is calculated, the retained pixels are selected, and a new low-frequency subband is created; S5, image fusion: the intensity image is fused with the low-frequency subband of the polarization image. The fusion weight is determined by the image gradient value, energy map information and empirical weight adjustment factor. The adaptive weight adjustment is completed by constructing the weight equation. S6, maximum fusion: perform maximum fusion on the high-frequency sub-bands and select the larger value of the pixels at the corresponding positions of the two images; S7, reconstruction: reconstruct the low-frequency sub-band and the high-frequency sub-band using the inverse transform of the image features to obtain a fused image.
3. The multi-scale image enhancement fusion method according to claim 2, characterized in that: In step S3, the intensity image and the polarization image are decomposed by discrete wavelet transform to obtain information in different frequency ranges; The information of different frequency ranges includes low frequency sub-band cA and high frequency sub-band. The high frequency sub-band includes horizontal high frequency sub-band cH, vertical high frequency sub-band cV and diagonal high frequency sub-band cD: Among them, cA j [k] represents the j-th level approximation subband, cD j [k] represents the j-th level detail subband, k represents the index of the subband, j represents the level of transformation, n represents the index variable in the transformation process, N represents the total length, ψ(·) represents the wavelet basis function, φ(·) represents the scaling function, and · represents the convolution operation.
4. The multi-scale image enhancement fusion method according to claim 3, characterized in that: In step S4, the formula for selecting the retained pixels is: Where I(q) represents the input image, G(p) represents the output image, q represents the image pixel coordinates of the input image, p represents the image pixel coordinates of the output image, σ s represents the standard deviation, σ r represents the spatial standard deviation of the control space range, N(p) represents the filter window centered at pixel p, and K p represents the normalization factor, J t (p) represents the result after the t-1th iteration, J t+1 (p) represents the result after the tth iteration.
5. The multi-scale image enhancement fusion method according to claim 4, characterized in that: Step S5 includes the following steps: S51, high frequency sub-band fusion: perform maximum value fusion on the horizontal high frequency sub-band cH, the vertical high frequency sub-band cV and the diagonal high frequency sub-band cD, and select the larger value of the pixels at the corresponding positions of the two images as the fusion result.
6. The multi-scale image enhancement fusion method according to claim 5, characterized in that: Step S5 includes the following steps: S52, low frequency sub-band fusion: the low frequency sub-bands are fused by weighted averaging.
7. The multi-scale image enhancement fusion method according to claim 6, characterized in that: In step S52, when the low-frequency sub-bands are fused, the calculation formula of the fusion weight is: α=Sig[ω]=Sig[(1-β)×G(i,j)+β×E(i,j)] Among them, α represents the fusion weight, ω represents the weight function, Sig[·] represents the mapping relationship, β represents the weight adjustment factor, (i, j) represents the pixel coordinates, G(i, j) represents the gradient map information of the image, and E(i, j)] represents the energy map information of the image.
8. The multi-scale image enhancement fusion method according to claim 2, characterized in that: In step S2, the intensity image is calculated separately to obtain a multi-parameter polarization image after the original polarization image is solved, and the parametric polarization image and the intensity image are selected as input images for image enhancement fusion.
9. A multi-scale image enhancement fusion method according to any one of claims 2 to 8, characterized in that: The following steps are also included: S7, equalization: histogram equalization: perform histogram equalization on the intensity image, polarization image, and fused image.
10. An enhanced fusion system, characterized in that: A multi-scale image enhancement fusion method for implementing any one of claims 1 to 9, comprising the following modules connected in sequence: Data acquisition module: used to build a polarization imaging detection system and simultaneously collect intensity images of multiple polarization angles in real scenes; Characterization and solution module: used to obtain a multi-parameter polarization image based on the intensity image characterization solution, select the polarization image and the intensity image for image alignment, and input the aligned image as an input image into the enhanced fusion system; Decomposition module: used to decompose the intensity image and the polarization image respectively using the frequency of the two-dimensional image features to obtain information in different frequency ranges, where the information in different frequency ranges includes low-frequency sub-bands and high-frequency sub-bands; Low-frequency sub-band creation module: used to calculate the weighted average of pixel values in the low-frequency sub-band by utilizing the edge preservation characteristic, select the retained pixels, and create a new low-frequency sub-band; Image fusion module: used to fuse the intensity image with the low-frequency subband of the polarization image. The fusion weight is determined by the image gradient value, energy map information and empirical weight adjustment factor. The adaptive weight adjustment is completed by constructing the weight equation. Maximum fusion module: used to perform maximum fusion on high-frequency sub-bands and select the larger value of pixels at corresponding positions of the two images; Reconstruction module: used to reconstruct the low-frequency sub-band and the high-frequency sub-band using the inverse transformation of image features to obtain a fused image.
