Local feature superposition and weighted fusion image enhancement method, device and equipment

Abstract

The invention discloses an image enhancement method, device and equipment based on local feature superposition and weighted fusion, and relates to the field of computer vision and image processing, an initial night vision image is segmented into a plurality of image sub-blocks, and local targeted processing is realized; performing illumination extraction operation of adaptive edge protection on each image sub-block, separating illumination components, accurately protecting edge information at the same time, and avoiding edge blurring; initial reflection component separation operation is carried out through a Retinex model, and an adaptive contrast adjustment factor is introduced for enhancement to obtain a target reflection component set, so that reflection component enhancement is more accurate; performing multi-scale wavelet decomposition operation on each image sub-block to obtain a low-frequency component and a weighted high-frequency component so as to realize targeted processing of the structure and details; and carrying out nonlinear fusion on the target reflection component set, the low-frequency component and the weighted high-frequency component of each image sub-block to obtain an enhanced night vision image, and improving local contrast while avoiding brightness imbalance under complex night vision illumination.

Description

Technical Field

[0001] This application relates to the fields of computer vision and image processing technology, and in particular to an image enhancement method, apparatus and device for local feature overlay and weighted fusion. Background Technology

[0002] With the continuous development of digital image processing technology, image enhancement, as a core task in computer vision and image processing, plays a vital role in many fields such as medical imaging, remote sensing monitoring, night vision imaging, and security monitoring. Image enhancement aims to improve the visual effect of images, enhance detail and contrast, thereby providing a higher-quality data foundation for subsequent image recognition, analysis, and understanding.

[0003] In related technologies, on the one hand, the Retinex algorithm, by decomposing an image into illumination and reflection components, has become a classic image enhancement method. The subsequently developed multi-scale Retinex (MSR) method, by introducing multi-scale illumination extraction, overcomes to some extent the limitations of single-scale Retinex in processing high dynamic range images. On the other hand, contrast-limited adaptive histogram equalization (CLAHE) significantly improves image contrast through local histogram adjustment, especially in unevenly lit scenes. However, the applicant recognizes that traditional Retinex algorithms are prone to producing halo artifacts in edge regions; while MSR alleviates this problem, it is sensitive to parameters and computationally complex; CLAHE is insufficient in enhancing details while improving overall contrast; and wavelet methods, although capable of enhancing details, lack effective modeling of illumination components, easily leading to brightness imbalances and artifacts under complex lighting conditions. Summary of the Invention

[0004] In view of this, this application provides an image enhancement method, apparatus and device that combines feature overlay and weighted fusion. The main purpose is to solve the problems of edge artifacts, parameter sensitivity, insufficient detail enhancement or lack of illumination modeling in existing image enhancement algorithms, which make it difficult to achieve high-quality enhancement in complex scenes.

[0005] According to a first aspect of this application, an image enhancement method based on feature overlay and weighted fusion is provided, the method comprising: Obtain an initial night view image and divide the initial night view image into multiple image sub-blocks; Perform adaptive edge-protected illumination extraction on each of the image sub-blocks to obtain the illumination component set of each image sub-block; The Retinex model is used to perform initial reflection component separation on the illumination component set of each image sub-block, and an adaptive contrast adjustment factor set for each image sub-block is introduced for enhancement, so as to obtain the target reflection component set of each image sub-block. Perform multi-scale wavelet decomposition on each of the image sub-blocks to obtain the low-frequency components and weighted high-frequency components of each of the image sub-blocks; The target reflection component set, low-frequency component, and weighted high-frequency component of each image sub-block are nonlinearly fused to obtain an enhanced night view image. The nonlinear fusion includes weighted superposition of the low-frequency component and the target reflection component set, and nonlinear transformation processing of the result of the weighted superposition and the weighted high-frequency component.

[0006] According to a second aspect of this application, an image enhancement apparatus for feature overlay and weighted fusion is provided, the apparatus comprising: The segmentation module is used to acquire an initial night view image and segment the initial night view image into multiple image sub-blocks; The illumination extraction module is used to perform adaptive edge-protected illumination extraction operations on each of the image sub-blocks to obtain the illumination component set of each of the image sub-blocks; The reflection enhancement module is used to perform initial reflection component separation operation on the illumination component set of each image sub-block using the Retinex model, and to enhance the target reflection component set of each image sub-block by introducing an adaptive contrast adjustment factor set of each image sub-block. The wavelet decomposition module is used to perform multi-scale wavelet decomposition on each of the image sub-blocks to obtain the low-frequency components and weighted high-frequency components of each of the image sub-blocks. The fusion module is used to perform nonlinear fusion of the target reflection component set, low-frequency component and weighted high-frequency component of each image sub-block to obtain an enhanced night view image. The nonlinear fusion includes weighted superposition of the low-frequency component and the target reflection component set, and nonlinear transformation processing of the result of the weighted superposition and the weighted high-frequency component.

[0007] According to a third aspect of this application, an apparatus is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the method described in any of the first aspects above.

[0008] By employing the above technical solutions, the technical solutions provided in the embodiments of this application have at least the following advantages: This application provides an image enhancement method, apparatus, and device that combines feature overlay and weighted fusion. The method involves acquiring an initial night view image, dividing it into multiple image sub-blocks, and enabling targeted local processing. Different sub-blocks can independently adjust enhancement parameters based on their own illumination distribution and detail density, avoiding over-enhancement or detail loss caused by global uniform processing. This approach is particularly suitable for complex images such as night view images and remote sensing images with uneven illumination and significant differences in detail distribution. Next, an adaptive edge-protected illumination extraction operation is performed on each image sub-block to obtain an illumination component set for each sub-block. An edge protection factor and adaptive standard deviation are introduced to accurately protect edge information while separating illumination, avoiding edge blurring and making the illumination extraction more closely match local features, thus improving the robustness of illumination separation. Subsequently, an initial reflection component separation operation is performed on the illumination component set of each image sub-block using a Retinex model, and an adaptive contrast adjustment factor set for each sub-block is introduced for enhancement, making the reflection component enhancement more precise, resulting in the target reflection component set for each image sub-block. Then, multi-scale wavelet decomposition is performed on each image sub-block to obtain the low-frequency component and weighted high-frequency component of each sub-block, achieving targeted processing of structure and detail. Finally, the target reflection component set, low-frequency component, and weighted high-frequency component of each image sub-block are nonlinearly fused to obtain the enhanced night view image. The nonlinear fusion includes weighted superposition of the low-frequency component and the target reflection component set, and nonlinear transformation processing is performed on the result of the weighted superposition and the weighted high-frequency component. The reflection component ensures basic details and contrast, the low-frequency component maintains the overall structural stability of the image, and the weighted high-frequency component enhances key details, avoiding structural distortion or detail overload caused by the dominance of a single component. The enhanced night view image retains the local adaptability of block processing and avoids obvious sub-block boundaries through edge smoothing, ensuring the global consistency and visual continuity of the enhanced image.

[0009] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0010] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This illustration shows a flowchart of an image enhancement method based on feature overlay and weighted fusion according to an embodiment of this application. Figure 2 This illustration shows a schematic flowchart of another image enhancement method based on feature overlay and weighted fusion provided in an embodiment of this application. Figure 3 This illustration shows a schematic diagram of an image enhancement structure based on feature overlay and weighted fusion according to an embodiment of this application. Figure 4 A schematic diagram of the device structure of an embodiment of this application is shown. Detailed Implementation

[0011] In the description of this application, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0012] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0013] In this application, unless otherwise expressly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0014] Exemplary embodiments of the present application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this application will be thorough and complete, and will fully convey the scope of the present application to those skilled in the art.

[0015] This application proposes an image enhancement method based on feature overlay and weighted fusion. First, the initial night view image is segmented into multiple image sub-blocks. Then, for each sub-block, adaptive edge-protected illumination extraction, Retinex+ adaptive contrast enhancement of the reflection component, and multi-scale wavelet decomposition are performed sequentially (separating low-frequency components reflecting the overall structure and high-frequency components reflecting details, with weighted enhancement of the high-frequency components). Finally, through nonlinear fusion, the enhanced reflection component, wavelet low-frequency component, and weighted high-frequency component are integrated to obtain the final enhanced image. The execution entity of this application can be an image enhancement system. The image enhancement system relies on the computing power of a server to provide services to users. The server can be a standalone server or a server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0016] This application provides an image enhancement method that combines feature overlay and weighted fusion, such as... Figure 1 As shown, the method includes: 101. Obtain the initial night view image and divide it into multiple image sub-blocks.

[0017] In this embodiment, an initial night view image to be enhanced is obtained, and then divided into several image sub-blocks according to a preset size. This decomposes global enhancement into local sub-block enhancement, allowing each sub-block to be processed specifically according to its own lighting and detail (such as edges and textures) features, avoiding uneven local enhancement caused by a uniform global operation. For example, when the night view image has mixed bright and dark areas, the local adaptability is stronger.

[0018] 102. Perform adaptive edge-protected illumination extraction on each image sub-block to obtain the illumination component set of each image sub-block.

[0019] In this embodiment, the mean local brightness and local contrast are calculated for each image sub-block, an edge protection factor is determined, and an adaptive Gaussian smoothing standard deviation that varies with local contrast is calculated. Finally, the image sub-block is convolved with a Gaussian function based on this standard deviation to extract the illumination components of each sub-block, forming an illumination component set. Compared to traditional fixed-parameter illumination extraction, the adaptive edge-protected illumination extraction operation can effectively separate illumination and accurately protect edges through the edge protection factor and adaptive standard deviation, thus better matching the local features of the sub-block.

[0020] 103. The initial reflection component separation operation is performed on the illumination component set of each image sub-block using the Retinex model, and the adaptive contrast adjustment factor set of each image sub-block is introduced for enhancement to obtain the target reflection component set of each image sub-block.

[0021] In this embodiment, the initial reflection component is separated by the Retinex model, and an adaptive contrast adjustment factor is introduced to enhance it, thereby obtaining the target reflection component set. The Retinex model separates the detail carrier at the physical level, which can solve the problem of light masking the details. The adaptive factor can fully enhance the low contrast area and avoid over-enhancing the high contrast area, thus balancing details and noise and improving local contrast.

[0022] 104. Perform multi-scale wavelet decomposition on each image sub-block to obtain the low-frequency component and weighted high-frequency component of each image sub-block.

[0023] In this embodiment, multi-scale wavelet decomposition is performed on each image sub-block to obtain low-frequency components reflecting the overall structure, as well as horizontal, vertical, and diagonal high-frequency components. Then, the high-frequency components are merged and weighted using a weighting factor based on local contrast to obtain weighted high-frequency components. Multi-scale wavelet decomposition adapts to details of different sizes, and the weighted high-frequency components can specifically enhance important details, suppress noise, and improve detail clarity.

[0024] 105. Nonlinearly fuse the target reflection component set, low-frequency component, and weighted high-frequency component of each image sub-block to obtain an enhanced night view image.

[0025] In this embodiment, the low-frequency component and the target reflection component set are weighted and superimposed. The result of the weighted superposition and the weighted high-frequency component are then subjected to nonlinear transformation. Instead of using a general fusion method with fixed weights, this method is based on the local features of image sub-blocks. It performs differentiated superposition and dynamic weighted fusion of the target reflection component set carrying pixel-level detail information, the low-frequency component carrying overall structure and illumination stability information, and the weighted high-frequency component carrying multi-scale texture detail information, driven by local features. Specifically, the low-frequency component mainly represents the macroscopic illumination and overall structure of the image, while the target reflection component set represents the inherent color and texture of the object. The weighted superposition is performed by assigning appropriate weights to both components and then fusing them. This is equivalent to initially superimposing and integrating the information representing illumination with the information representing the object itself within the local area of ​​each image sub-block to establish an image foundation with good contrast and balanced illumination. The weighted high-frequency component contains weighted edge and texture detail information. The nonlinear transformation differs from simple linear mixing; the nonlinear function can significantly enhance strong details in a more intelligent way while suppressing weak noise, thereby "injecting" detail information into the base image generated in the first step in a controlled and nonlinear manner. Therefore, nonlinear fusion based on the target reflection component set, low-frequency component and weighted high-frequency component integrates the advantages of each component, balances the contribution of structure, detail and contrast, avoids distortion or overload caused by the dominance of a single component, and finally obtains a visually natural and detailed enhancement result.

[0026] This application provides an image enhancement method based on feature overlay and weighted fusion. Compared with existing technologies, this application acquires an initial night view image and divides it into multiple image sub-blocks. By dividing the image into multiple sub-blocks, targeted local processing is achieved. Different sub-blocks can independently adjust enhancement parameters according to their own illumination distribution and detail density, avoiding local over-enhancement or detail loss caused by global uniform processing. This method is particularly suitable for complex images such as night view images and remote sensing images with uneven illumination and large differences in detail distribution. Next, an adaptive edge-protected illumination extraction operation is performed on each image sub-block to obtain the illumination component set of each sub-block. An edge protection factor and adaptive standard deviation are introduced to accurately protect edge information while separating illumination, avoiding edge blurring and making the illumination extraction more closely match local features, thus improving the robustness of illumination separation. Subsequently, an initial reflection component separation operation is performed on the illumination component set of each image sub-block using a Retinex model, and an adaptive contrast adjustment factor set for each image sub-block is introduced for enhancement, making the reflection component enhancement more accurate, resulting in the target reflection component set for each image sub-block. Then, multi-scale wavelet decomposition is performed on each image sub-block to obtain the low-frequency component and weighted high-frequency component of each sub-block, achieving targeted processing of structure and detail. Finally, the target reflection component set, low-frequency component, and weighted high-frequency component of each image sub-block are nonlinearly fused to obtain the enhanced night view image. The nonlinear fusion includes weighted superposition of the low-frequency component and the target reflection component set, and nonlinear transformation processing is performed on the result of the weighted superposition and the weighted high-frequency component. The reflection component ensures basic details and contrast, the low-frequency component maintains the overall structural stability of the image, and the weighted high-frequency component enhances key details, avoiding structural distortion or detail overload caused by the dominance of a single component. The enhanced night view image retains the local adaptability of block processing and avoids obvious sub-block boundaries through edge smoothing, ensuring the global consistency and visual continuity of the enhanced image.

[0027] Furthermore, as a refinement and extension of the specific implementation methods of the above embodiments, in order to fully illustrate the specific implementation process of this embodiment, this application provides another image enhancement method based on feature overlay and weighted fusion, such as... Figure 2 As shown, the method includes: 201. Obtain the initial night view image and divide it into multiple image sub-blocks.

[0028] In this embodiment, the image sub-block segmentation size is determined, and the initial night view image is divided into multiple image sub-blocks according to the image sub-block segmentation size, which facilitates the processing of local features. Assume the initial night view image size is... Divide it into Image sub-blocks are used to enhance the local adaptability of the image, as shown in Formula 1 below: Formula 1:

[0029] in, This represents the initial night view image. This represents the r-th image sub-block. This indicates the total number of image sub-blocks.

[0030] It should be noted that the input image can be night vision images used in scenarios such as vehicle-mounted and security systems, or ordinary natural images with low contrast or blurred details. The application areas of this application cover vehicle-mounted systems, security monitoring (night vision), and image preprocessing in computer vision (preparing for subsequent recognition, analysis, and other tasks).

[0031] 202. Perform adaptive edge-protected illumination extraction on each image sub-block to obtain the illumination component set of each image sub-block.

[0032] In this embodiment of the application, for each image sub-block, the local brightness mean set of the image sub-block is calculated as shown in Formula 2 below: Formula 2:

[0033] in, Represents the image sub-block at the target pixel coordinates The local average brightness at that location, Indicates the image sub-block at offset pixel coordinates Pixel value at that location, Indicates that it contains the target pixel coordinates The neighborhood refers to the area of ​​the neighborhood of the neighboring country. The neighborhood is a local region centered on the i-th image sub-block, and is not necessarily limited to the set of pixels within the i-th image sub-block. The specific range can be determined according to the neighborhood size set by the algorithm, and may span across sub-blocks or be within sub-blocks. This represents the image size of the sub-block. It should be noted that i and j represent the pixel coordinates within the sub-block, i.e., the row and column indices. and Both are used to represent the coordinates of a pixel. Here are the coordinates of the target pixel whose local brightness average is to be calculated. For the neighborhood The pixel coordinates used in the summation calculation are the offset pixel coordinates.

[0034] Next, the local contrast set of the image sub-blocks is calculated using the local brightness mean set of the image sub-blocks, providing a basis for subsequently determining the adaptive standard deviation, as shown in Formula 3 below: Formula 3:

[0035] in, Represents the image sub-block at the target pixel coordinates Local contrast at that location Represents the image sub-block at the target pixel coordinates The local average brightness at that location, Indicates the image sub-block at offset pixel coordinates Pixel value at that location, Indicates that it contains the target pixel coordinates The neighborhood, This represents the image size of a sub-block. Local brightness mean and local contrast are used to enhance local details of the image using statistical information. By calculating pixel-level local brightness mean and local contrast, the brightness distribution and detail differences around each pixel are accurately characterized in a neighborhood statistical manner. This provides fine-grained local information for subsequent adaptive adjustment, overcoming the limitations of traditional globally uniform parameters and making it more suitable for scenes with uneven lighting and large fluctuations in detail density.

[0036] Determine multiple gradient values ​​for the image sub-block, and use these gradient values ​​to calculate the edge protection factor set for the image sub-block, as shown in Formula 4 below: Formula 4:

[0037] in, Represents the image sub-block at the target pixel coordinates Edge protection factor reduces the smoothing effect on edge regions by suppressing the gradient of the image. Represents the image sub-block at the target pixel coordinates gradient value at, This parameter represents the control gradient's effect on the edge protection factor. Closer to 1, it has a stronger inhibitory effect on Gaussian smoothing, reducing over-smoothing of edges and accurately preserving edge contours; the gradient in the smoothed region is small. Smaller size allows for stronger smoothing, effectively uniform lighting and suppressing noise, thus achieving an initial balance between edge preservation and lighting uniformity.

[0038] An adaptive standard deviation set for image sub-blocks is calculated using the edge protection factor set and local contrast set of the image sub-blocks to achieve a dynamic smoothing effect, as shown in Formula 5 below: Formula 5:

[0039] in, Represents the image sub-block at the target pixel coordinates Adaptive standard deviation at the location, Represents the image sub-block at the target pixel coordinates Local contrast at that location Represents the image sub-block at the target pixel coordinates Edge protection factor at the location, Represents the initial standard deviation constant. This parameter controls the impact of contrast on the standard deviation. Adaptive standard deviation integrates local contrast and edge protection factors, dynamically matching the smoothing intensity of each pixel to its local features.

[0040] Gaussian smoothing is applied to the adaptive standard deviation set of the image sub-blocks to obtain the illumination component set of the image sub-blocks, as shown in Formula 6 below: Formula 6:

[0041] in, Represents the image sub-block at the target pixel coordinates The amount of light at that location, Represents the image sub-block at the target pixel coordinates Pixel value at that location, Represents the image sub-block at the target pixel coordinates Adaptive standard deviation at the location, The standard deviation is expressed as The Gaussian function.

[0042] 203. Using the Retinex model, perform initial reflection component separation on the illumination component set of each image sub-block to obtain the initial reflection component set of each image sub-block.

[0043] In this embodiment, Retinex theory is used to separate the reflection components, and adaptive contrast enhancement is used to improve image details. For each image sub-block, the initial reflection component separation operation is performed on the illumination component set of the image sub-block using the Retinex model to obtain the initial reflection component set of the image sub-block, as shown in Formula 7 below: Formula 7:

[0044] in, Represents the image sub-block at the target pixel coordinates The initial reflection component at that location, This represents a very small constant value, used to prevent logarithmic numerical overflow. Represents the image sub-block at the target pixel coordinates Pixel value at that location, Represents the image sub-block at the target pixel coordinates The amount of light at that location.

[0045] 204. Introduce an adaptive contrast adjustment factor set for each image sub-block to enhance the initial reflection component set of each image sub-block, thereby obtaining the target reflection component set of each image sub-block.

[0046] In this embodiment, the global maximum contrast and local contrast set of the image sub-block are determined, and the adaptive contrast adjustment factor set of the image sub-block is calculated using the global maximum contrast and local contrast set of the image sub-block to achieve pixel-level differential enhancement, as shown in the following formula 8: Formula 8:

[0047] in, Represents the image sub-block at the target pixel coordinates Adaptive contrast adjustment factor at the location, Represents the image sub-block at the target pixel coordinates Local contrast at that location This represents the global maximum contrast of a sub-block of the image. This represents a very small constant value.

[0048] The initial reflection component set of an image sub-block is enhanced using an adaptive contrast adjustment factor set, resulting in an enhanced reflection component set and thus higher contrast, as shown in Formula 9 below: Formula 9:

[0049] in, Represents the image sub-block at the target pixel coordinates Enhanced reflection component at the location, Represents the image sub-block at the target pixel coordinates The initial reflection component at that location, Represents the image sub-block at the target pixel coordinates The adaptive contrast adjustment factor at that location.

[0050] Obtain the nonlinear adjustment parameters, and use the nonlinear adjustment parameters to perform nonlinear mapping on the enhanced reflectance component set of the image sub-block to improve brightness perception, thus obtaining the target reflectance component set of the image sub-block, as shown in Formula 10 below: Formula 10:

[0051] in, Represents the image sub-block at the target pixel coordinates The target reflection component at that location, Represents the image sub-block at the target pixel coordinates Adaptive contrast adjustment factor at the location, This represents a very small constant value. Represents the image sub-block at the target pixel coordinates Pixel value at that location, Represents the image sub-block at the target pixel coordinates The amount of light at that location, This represents the nonlinear adjustment parameter.

[0052] 205. Perform multi-scale wavelet decomposition on each image sub-block to obtain the low-frequency component and weighted high-frequency component of each image sub-block.

[0053] In this embodiment, low-frequency and high-frequency components are extracted by wavelet decomposition, and detail-weighted processing is applied to the high-frequency components. For each image sub-block, wavelet decomposition is performed to obtain the wavelet decomposition result of the image sub-block, avoiding interference from global processing on structure or details, as shown in Formula 11 below: Formula 11:

[0054] in, This represents the wavelet decomposition result of the r-th image sub-block. This represents the r-th image sub-block. This represents the low-frequency component of the r-th image sub-block. This represents the horizontal high-frequency component of the r-th image sub-block. This represents the high-frequency component in the vertical direction of the r-th image sub-block. This represents the high-frequency component in the diagonal direction of the r-th image sub-block.

[0055] The high-frequency components in the horizontal, vertical, and diagonal directions of the wavelet decomposition results of the image sub-blocks are merged to obtain the merged high-frequency components of the image sub-blocks. These merged components are used to enhance local details, enabling subsequent enhancement operations to simultaneously cover details in all directions. This avoids the problem of enhancing details in only one dimension while losing details in other dimensions, thus improving the integrity and comprehensiveness of the details, as shown in Formula 12 below. Formula 12:

[0056] in, This represents the merged high-frequency components of the r-th image sub-block. This represents the horizontal high-frequency component of the r-th image sub-block. This represents the high-frequency component in the vertical direction of the r-th image sub-block. This represents the high-frequency component in the diagonal direction of the r-th image sub-block.

[0057] The detail weighting factor set of the image sub-block is determined by using the local brightness mean set and local contrast set of the image sub-block. The detail weighting factor matrix of the image sub-block is constructed using the detail weighting factor set, which is used to weight and enhance high-frequency components, avoiding noise amplification or excessive enhancement of invalid information, and ensuring that the detail enhancement accurately matches the local features, as shown in Formula 13 below: Formula 13:

[0058] in, This represents the target pixel coordinates of the r-th image sub-block. The detail weighting factor, This represents the target pixel coordinates of the r-th image sub-block. The local average brightness at that location, This represents the target pixel coordinates of the r-th image sub-block. Local contrast at that location This represents a very small constant value.

[0059] The image details are enhanced by using a detail weighting factor matrix for image sub-blocks and merging high-frequency components to determine the weighted high-frequency components of the image sub-blocks, as shown in Formula 14 below: Formula 14:

[0060] in, This represents the weighted high-frequency component of the r-th image sub-block. This represents the detail weighting factor matrix of the r-th image sub-block. This represents the merged high-frequency components of the r-th image sub-block.

[0061] 206. Nonlinearly fuse the target reflection component set, low-frequency component, and weighted high-frequency component of each image sub-block to obtain an enhanced night view image.

[0062] In this embodiment of the application, for each image sub-block, the target reflection component set, low-frequency component, and weighted high-frequency component of the image sub-block are nonlinearly fused to obtain an enhanced image sub-block, so as to achieve a balance between detail and brightness, as shown in the following formula 15: Formula 15:

[0063] in, This represents the target pixel coordinates of the r-th enhanced image sub-block. Pixel value at that location, This represents the target pixel coordinates of the r-th image sub-block. The target reflection component at that location, This represents the weighted high-frequency component of the r-th image sub-block. This represents the low-frequency component of the r-th image sub-block. This represents a very small constant value. Indicates the first fusion index, This indicates the second fusion index. It represents the exponent of the power transformation.

[0064] The enhanced image sub-blocks obtained by nonlinearly fusing each image sub-block are obtained, resulting in multiple enhanced image sub-blocks. The multiple enhanced image sub-blocks are then recombined according to the target pixel coordinates of each enhanced image sub-block to obtain an enhanced night view image, as shown in Formula 16 below: Formula 16:

[0065] in, This indicates enhanced night view images. This represents the r-th enhanced image sub-block. This indicates the total number of enhanced image sub-blocks.

[0066] This application integrates an improved Retinex method with wavelet decomposition techniques, constructing a precise environment perception and detail enhancement mechanism through illumination-adaptive component extraction, nonlinear enhancement of reflection components, and multi-scale detail weighted fusion. This mechanism aims to more effectively improve image contrast, detail, and brightness, reducing artifacts while enhancing visual effects.

[0067] To verify the effectiveness of this application, experiments were conducted on multiple sets of images, and the results were compared with classic multi-scale Retinex (MSR) and CLAHE algorithms. The enhancement effects of each algorithm were evaluated using objective metrics, including peak signal-to-noise ratio (PSNR), structural similarity (SSIM), and contrast enhancement.

[0068] The single-scale Retinex algorithm is one of the early methods in the field of image enhancement. Its basic idea is to decompose an image into illumination and reflection components to improve image contrast. However, due to the use of a fixed-scale illumination component separation method, single-scale Retinex has limited effectiveness in handling complex lighting conditions (such as uneven illumination). It often produces halo artifacts at high-contrast edges and has weak detail enhancement capabilities, thus limiting its application in complex scenes.

[0069] Multi-scale Retinex (MSR) is an improvement over single-scale Retinex, employing multiple scales to decompose illumination components to adapt to a wider range of lighting conditions. Through multi-scale processing, MSR achieves better contrast and detail enhancement under varying illumination intensities, making it suitable for images with high dynamic range. While MSR effectively improves detail enhancement, its computational complexity is high, and it is highly sensitive to parameter selection. During enhancement, although MSR reduces halo effects, some loss of edge detail still occurs.

[0070] CLAHE is a histogram equalization-based enhancement method that avoids over-enhancing bright or dark areas by limiting the increase in local contrast. CLAHE performs well in scenes with significant lighting variations, effectively enhancing image contrast without oversaturation. However, CLAHE has limited detail enhancement, especially at image edges and complex textures. Its enhancement effect mainly relies on the image's histogram distribution, offering less direct detail enhancement and lacking in brightness information preservation.

[0071] Image enhancement methods based on wavelet transform decompose the low-frequency and high-frequency components of an image, processing smooth and detailed regions separately to achieve enhancement. Wavelet enhancement possesses good localization properties in both the frequency and spatial domains, making it suitable for enhancing details and edge information, enhancing local details while preserving the overall structure. However, wavelet transform primarily focuses on frequency component decomposition and does not separate the illumination and reflection components of the image. Therefore, it is prone to insufficient brightness and contrast in uneven or complex lighting environments. Compared to the Retinex method, wavelet enhancement is less effective at adjusting the overall brightness of the image and is more likely to produce artifacts in areas of uneven brightness.

[0072] Table 1

[0073] Table 1 shows the algorithm comparison results, which demonstrate that our application outperforms the compared algorithms in PSNR, SSIM, and contrast enhancement rate. Our application effectively reduces noise amplification and detail loss, and improves the signal-to-noise ratio through adaptive illumination extraction and wavelet weighted fusion. Compared to MSR and CLAHE, our application preserves image structure better and enhances details more naturally. Furthermore, our application performs best in image contrast enhancement, thanks to the nonlinear enhancement of the reflection component and detail weighting processing.

[0074] This application combines the advantages of Retinex theory and wavelet transform, achieving a balance between detail preservation, edge retention, and brightness adjustment through adaptive edge-protected illumination extraction, nonlinear reflection component enhancement, and multi-scale wavelet detail fusion. Adaptive processing of the illumination component using an edge protection factor ensures that enhanced contrast does not damage edge details; adaptive nonlinear reflection component adjustment effectively improves local contrast, resulting in more natural detail enhancement; and multi-scale wavelet fusion balances low-frequency and high-frequency components during enhancement, thereby enhancing the overall contrast and detail levels of the image. Compared to contrast-based methods, this application performs particularly well in images with complex lighting and high dynamic range, reducing halo artifacts while enhancing detail levels, resulting in a visually clearer and more natural final effect.

[0075] This application provides an image enhancement method based on feature overlay and weighted fusion. Compared with existing technologies, this application acquires an initial night view image and divides it into multiple image sub-blocks. By dividing the image into multiple sub-blocks, targeted local processing is achieved. Different sub-blocks can independently adjust enhancement parameters according to their own illumination distribution and detail density, avoiding local over-enhancement or detail loss caused by global uniform processing. This method is particularly suitable for complex images such as night view images and remote sensing images with uneven illumination and large differences in detail distribution. Next, an adaptive edge-protected illumination extraction operation is performed on each image sub-block to obtain the illumination component set of each sub-block. An edge protection factor and adaptive standard deviation are introduced to accurately protect edge information while separating illumination, avoiding edge blurring and making the illumination extraction more closely match local features, thus improving the robustness of illumination separation. Subsequently, an initial reflection component separation operation is performed on the illumination component set of each image sub-block using a Retinex model, and an adaptive contrast adjustment factor set for each image sub-block is introduced for enhancement, making the reflection component enhancement more accurate, resulting in the target reflection component set for each image sub-block. Then, multi-scale wavelet decomposition is performed on each image sub-block to obtain the low-frequency component and weighted high-frequency component of each sub-block, achieving targeted processing of structure and detail. Finally, the target reflection component set, low-frequency component, and weighted high-frequency component of each image sub-block are nonlinearly fused to obtain the enhanced night view image. The nonlinear fusion includes weighted superposition of the low-frequency component and the target reflection component set, and nonlinear transformation processing is performed on the result of the weighted superposition and the weighted high-frequency component. The reflection component ensures basic details and contrast, the low-frequency component maintains the overall structural stability of the image, and the weighted high-frequency component enhances key details, avoiding structural distortion or detail overload caused by the dominance of a single component. The enhanced night view image retains the local adaptability of block processing and avoids obvious sub-block boundaries through edge smoothing, ensuring the global consistency and visual continuity of the enhanced image.

[0076] Furthermore, as Figure 1 In a specific implementation of the method, this application provides an image enhancement device that combines feature overlay and weighted fusion, such as... Figure 3 As shown, the device includes: a segmentation module 301, an illumination extraction module 302, a reflection enhancement module 303, a wavelet decomposition module 304, and a fusion module 305.

[0077] The segmentation module 301 is used to acquire an initial night view image and segment the initial night view image into multiple image sub-blocks; The illumination extraction module 302 is used to perform adaptive edge-protected illumination extraction operations on each of the image sub-blocks to obtain the illumination component set of each of the image sub-blocks; The reflection enhancement module 303 is used to perform initial reflection component separation operation on the illumination component set of each image sub-block using the Retinex model, and to enhance the target reflection component set of each image sub-block by introducing an adaptive contrast adjustment factor set of each image sub-block. Wavelet decomposition module 304 is used to perform multi-scale wavelet decomposition operation on each of the image sub-blocks to obtain the low-frequency component and weighted high-frequency component of each of the image sub-blocks. The fusion module 305 is used to perform nonlinear fusion on the target reflection component set, low frequency component and weighted high frequency component of each image sub-block to obtain an enhanced night view image. The nonlinear fusion includes weighted superposition of the low frequency component and the target reflection component set, and nonlinear transformation processing of the result of the weighted superposition and the weighted high frequency component.

[0078] In specific application scenarios, the segmentation module 301 is used to determine the image sub-block segmentation size and segment the initial night view image into multiple image sub-blocks according to the image sub-block segmentation size.

[0079] in, This represents the initial night view image. This represents the r-th image sub-block. This indicates the total number of image sub-blocks.

[0080] In specific application scenarios, the illumination extraction module 302 is used to calculate the local brightness mean set of each image sub-block.

[0081] in, Indicates the image sub-block at the target pixel coordinates The local average brightness at that location, This indicates that the image sub-block is at offset pixel coordinates. Pixel value at that location, Indicates that it contains the target pixel coordinates The neighborhood, The image size of the image sub-block is represented; the local contrast set of the image sub-block is calculated using the local brightness mean set of the image sub-block.

[0082] in, Indicates the image sub-block at the target pixel coordinates Local contrast at that location Indicates the image sub-block at the target pixel coordinates The local average brightness at that location, This indicates that the image sub-block is at offset pixel coordinates. Pixel value at that location, Indicates that it contains the target pixel coordinates The neighborhood, The image size of the image sub-block is represented; multiple gradient values ​​of the image sub-block are determined, and the edge protection factor set of the image sub-block is calculated using the multiple gradient values ​​of the image sub-block.

[0083] in, Indicates the image sub-block at the target pixel coordinates Edge protection factor at the location, Indicates the image sub-block at the target pixel coordinates gradient value at, The parameter represents the influence of the control gradient on the edge protection factor; the adaptive standard deviation set of the image sub-block is calculated using the edge protection factor set and the local contrast set of the image sub-block.

[0084] in, Indicates the image sub-block at the target pixel coordinates Adaptive standard deviation at the location, Indicates the image sub-block at the target pixel coordinates Local contrast at that location Indicates the image sub-block at the target pixel coordinates Edge protection factor at the location, Represents the initial standard deviation constant. The parameter represents the effect of contrast on the standard deviation; Gaussian smoothing is performed on the adaptive standard deviation set of the image sub-block to obtain the illumination component set of the image sub-block.

[0085] in, Indicates the image sub-block at the target pixel coordinates The amount of light at that location, Indicates the image sub-block at the target pixel coordinates Pixel value at that location, Indicates the image sub-block at the target pixel coordinates Adaptive standard deviation at the location, The standard deviation is expressed as The Gaussian function.

[0086] In specific application scenarios, the reflection enhancement module 303 is used to perform initial reflection component separation on the illumination component set of each image sub-block using the Retinex model, thereby obtaining the initial reflection component set of the image sub-block.

[0087] in, Indicates the image sub-block at the target pixel coordinates The initial reflection component at that location, This represents a very small constant value. Indicates the image sub-block at the target pixel coordinates Pixel value at that location, Indicates the image sub-block at the target pixel coordinates The illumination component at the location is determined; the adaptive contrast adjustment factor set of the image sub-block is determined, and the initial reflection component set of the image sub-block is enhanced using the adaptive contrast adjustment factor set of the image sub-block to obtain the target reflection component set of the image sub-block.

[0088] In specific application scenarios, the reflection enhancement module 303 is used to determine the global maximum contrast and local contrast set of the image sub-block, and to calculate the adaptive contrast adjustment factor set of the image sub-block using the global maximum contrast and local contrast set of the image sub-block.

[0089] in, Indicates the image sub-block at the target pixel coordinates Adaptive contrast adjustment factor at the location, Indicates the image sub-block at the target pixel coordinates Local contrast at that location This represents the global maximum contrast of the image sub-block. This represents a very small constant value; the initial reflection component set of the image sub-block is enhanced using the adaptive contrast adjustment factor set of the image sub-block to obtain the enhanced reflection component set of the image sub-block.

[0090] in, Indicates the image sub-block at the target pixel coordinates Enhanced reflection component at the location, Indicates the image sub-block at the target pixel coordinates The initial reflection component at that location, Indicates the image sub-block at the target pixel coordinates An adaptive contrast adjustment factor is determined at the specified location; a nonlinear adjustment parameter is obtained, and the enhanced reflectance component set of the image sub-block is nonlinearly mapped using the nonlinear adjustment parameter to obtain the target reflectance component set of the image sub-block.

[0091] in, Indicates the image sub-block at the target pixel coordinates The target reflection component at that location, Indicates the image sub-block at the target pixel coordinates Adaptive contrast adjustment factor at the location, This represents a very small constant value. Indicates the image sub-block at the target pixel coordinates Pixel value at that location, Indicates the image sub-block at the target pixel coordinates The amount of light at that location, This represents the nonlinear adjustment parameter.

[0092] In specific application scenarios, the wavelet decomposition module 304 is used to perform wavelet decomposition on each of the image sub-blocks to obtain the wavelet decomposition result of the image sub-block.

[0093] in, This represents the wavelet decomposition result of the r-th image sub-block. This represents the r-th image sub-block. This represents the low-frequency component of the r-th image sub-block. This represents the horizontal high-frequency component of the r-th image sub-block. This represents the high-frequency component in the vertical direction of the r-th image sub-block. Let represent the diagonal high-frequency component of the r-th image sub-block; the horizontal, vertical, and diagonal high-frequency components in the wavelet decomposition result of the image sub-block are merged to obtain the merged high-frequency component of the image sub-block.

[0094] in, This represents the merged high-frequency components of the r-th image sub-block. This represents the horizontal high-frequency component of the r-th image sub-block. This represents the high-frequency component in the vertical direction of the r-th image sub-block. Let represent the high-frequency components in the diagonal direction of the r-th image sub-block; determine the detail weighting factor set of the image sub-block using the local brightness mean set and local contrast set of the image sub-block; and construct the detail weighting factor matrix of the image sub-block using the detail weighting factor set.

[0095] in, This represents the target pixel coordinates of the r-th image sub-block. The detail weighting factor, This represents the target pixel coordinates of the r-th image sub-block. The local average brightness at that location, This represents the target pixel coordinates of the r-th image sub-block. Local contrast at that location This represents a very small constant value; the weighted high-frequency components of the image sub-block are determined using the detail weighting factor matrix of the image sub-block and by merging the high-frequency components.

[0096] in, This represents the weighted high-frequency component of the r-th image sub-block. This represents the detail weighting factor matrix of the r-th image sub-block. This represents the merged high-frequency components of the r-th image sub-block.

[0097] In specific application scenarios, the fusion module 305 is used to perform nonlinear fusion of the target reflection component set, low-frequency component, and weighted high-frequency component of each image sub-block to obtain an enhanced image sub-block.

[0098] in, This represents the target pixel coordinates of the r-th enhanced image sub-block. Pixel value at that location, This represents the target pixel coordinates of the r-th image sub-block. The target reflection component at that location, This represents the weighted high-frequency component of the r-th image sub-block. This represents the low-frequency component of the r-th image sub-block. This represents a very small constant value. Indicates the first fusion index, This indicates the second fusion index. The power transformation exponent is represented; the enhanced image sub-blocks obtained by nonlinearly fusing each of the image sub-blocks are obtained, resulting in multiple enhanced image sub-blocks; the multiple enhanced image sub-blocks are recombined to obtain the enhanced night view image.

[0099] In a specific application scenario, the fusion module 305 is used to reassemble the plurality of enhanced image sub-blocks according to the target pixel coordinates of each enhanced image sub-block to obtain the enhanced night view image.

[0100] in, This refers to the enhanced night view image. This represents the r-th enhanced image sub-block. This indicates the total number of enhanced image sub-blocks.

[0101] This application provides an apparatus that, compared to existing technologies, acquires an initial night view image and divides it into multiple image sub-blocks. By dividing the image into multiple sub-blocks, targeted local processing is achieved. Different sub-blocks can independently adjust enhancement parameters according to their own illumination distribution and detail density, avoiding local over-enhancement or detail loss caused by global uniform processing. This is particularly suitable for complex images such as night view images and remote sensing images with uneven illumination and large differences in detail distribution. Next, an adaptive edge-protected illumination extraction operation is performed on each image sub-block to obtain the illumination component set of each sub-block. An edge protection factor and adaptive standard deviation are introduced to accurately protect edge information while separating illumination, avoiding edge blurring, and making the illumination extraction more closely match local features, thus improving the robustness of illumination separation. Subsequently, an initial reflection component separation operation is performed on the illumination component set of each image sub-block using a Retinex model, and an adaptive contrast adjustment factor set for each image sub-block is introduced for enhancement, making the reflection component enhancement more accurate, resulting in the target reflection component set of each image sub-block. Then, multi-scale wavelet decomposition is performed on each image sub-block to obtain the low-frequency component and weighted high-frequency component of each sub-block, achieving targeted processing of structure and detail. Finally, the target reflection component set, low-frequency component, and weighted high-frequency component of each image sub-block are nonlinearly fused to obtain the enhanced night view image. The nonlinear fusion includes weighted superposition of the low-frequency component and the target reflection component set, and nonlinear transformation processing is performed on the result of the weighted superposition and the weighted high-frequency component. The reflection component ensures basic details and contrast, the low-frequency component maintains the overall structural stability of the image, and the weighted high-frequency component enhances key details, avoiding structural distortion or detail overload caused by the dominance of a single component. The enhanced night view image retains the local adaptability of block processing and avoids obvious sub-block boundaries through edge smoothing, ensuring the global consistency and visual continuity of the enhanced image.

[0102] It should be noted that other corresponding descriptions of the functional units involved in the image enhancement device for feature overlay and weighted fusion provided in this application embodiment can be found in the following references. Figure 1 and Figure 2 The corresponding descriptions in [the document] will not be repeated here.

[0103] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0104] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0105] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

[0106] In an exemplary embodiment, see Figure 4 Furthermore, a device is provided, comprising a bus, a processor, a memory, and a communication interface. It may also include an input / output interface and a display device, wherein the various functional units can communicate with each other via the bus. The memory stores a computer program, and the processor executes the program stored in the memory to perform the image enhancement method of feature overlay and weighted fusion described in the above embodiments.

[0107] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the image enhancement method of feature overlay and weighted fusion.

[0108] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented in hardware or by using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) and includes several instructions to cause a computer device (such as a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0109] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing this application.

[0110] Those skilled in the art will understand that the modules in the apparatus of the implementation scenario can be distributed within the apparatus of the implementation scenario as described, or they can be located in one or more apparatuses different from this implementation scenario, with corresponding changes. The modules of the above-described implementation scenario can be combined into one module, or they can be further divided into multiple sub-modules.

[0111] The serial numbers in this application are for descriptive purposes only and do not represent the superiority or inferiority of the implementation scenario.

[0112] The above disclosures are only a few specific implementation scenarios of this application. However, this application is not limited to these. Any variations that can be conceived by those skilled in the art should fall within the protection scope of this application.

Claims

1. An image enhancement method based on local feature overlay and weighted fusion, characterized in that, include: Obtain an initial night view image and divide the initial night view image into multiple image sub-blocks; Perform adaptive edge-protected illumination extraction on each of the image sub-blocks to obtain the illumination component set of each image sub-block; The Retinex model is used to perform initial reflection component separation on the illumination component set of each image sub-block, and an adaptive contrast adjustment factor set for each image sub-block is introduced for enhancement, so as to obtain the target reflection component set of each image sub-block. Perform multi-scale wavelet decomposition on each of the image sub-blocks to obtain the low-frequency components and weighted high-frequency components of each of the image sub-blocks; The target reflection component set, low-frequency component, and weighted high-frequency component of each image sub-block are nonlinearly fused to obtain an enhanced night view image. The nonlinear fusion includes weighted superposition of the low-frequency component and the target reflection component set, and nonlinear transformation processing of the result of the weighted superposition and the weighted high-frequency component.

2. The method according to claim 1, characterized in that, The step of dividing the initial night view image into multiple image sub-blocks includes: Determine the image sub-block segmentation size, and divide the initial night view image into multiple image sub-blocks according to the determined image sub-block segmentation size. in, This represents the initial night view image. This represents the r-th image sub-block. This indicates the total number of image sub-blocks.

3. The method according to claim 1, characterized in that, The step of performing adaptive edge-protected illumination extraction on each of the image sub-blocks to obtain the illumination component set of each image sub-block includes: For each of the image sub-blocks, calculate the local brightness mean set of the image sub-block. in, Indicates the image sub-block at the target pixel coordinates The local average brightness at that location, This indicates that the image sub-block is at offset pixel coordinates. Pixel value at that location, Indicates that it contains the target pixel coordinates The neighborhood, This indicates the image size of the image sub-block; The local contrast set of the image sub-block is calculated using the local brightness mean set of the image sub-block. in, Indicates the image sub-block at the target pixel coordinates Local contrast at that location Indicates the image sub-block at the target pixel coordinates The local average brightness at that location, This indicates that the image sub-block is at offset pixel coordinates. Pixel value at that location, Indicates that it contains the target pixel coordinates The neighborhood, This indicates the image size of the image sub-block; Multiple gradient values ​​of the image sub-block are determined, and the edge protection factor set of the image sub-block is calculated using the multiple gradient values ​​of the image sub-block. in, Indicates the image sub-block at the target pixel coordinates Edge protection factor at the location, Indicates the image sub-block at the target pixel coordinates gradient value at, The parameter represents the control gradient's effect on the edge protection factor; The adaptive standard deviation set of the image sub-block is calculated using the edge protection factor set and local contrast set of the image sub-block. in, Indicates the image sub-block at the target pixel coordinates Adaptive standard deviation at the location, Indicates the image sub-block at the target pixel coordinates Local contrast at that location Indicates the image sub-block at the target pixel coordinates Edge protection factor at the location, Represents the initial standard deviation constant. This parameter represents the effect of controlling contrast on the standard deviation. Gaussian smoothing is performed on the adaptive standard deviation set of the image sub-block to obtain the illumination component set of the image sub-block. in, Indicates the image sub-block at the target pixel coordinates The amount of light at that location, Indicates the image sub-block at the target pixel coordinates Pixel value at that location, Indicates the image sub-block at the target pixel coordinates Adaptive standard deviation at the location, The standard deviation is expressed as The Gaussian function.

4. The method according to claim 1, characterized in that, The process involves performing initial reflectance component separation on the illumination component set of each image sub-block using the Retinex model, and then enhancing it by introducing an adaptive contrast adjustment factor set for each image sub-block to obtain the target reflectance component set for each image sub-block, including: For each image sub-block, the initial reflection component separation operation is performed on the illumination component set of the image sub-block using the Retinex model to obtain the initial reflection component set of the image sub-block. in, Indicates the image sub-block at the target pixel coordinates The initial reflection component at that location, This represents a very small constant value. Indicates the image sub-block at the target pixel coordinates Pixel value at that location, Indicates the image sub-block at the target pixel coordinates The amount of light at that location; Determine the adaptive contrast adjustment factor set of the image sub-block, and use the adaptive contrast adjustment factor set of the image sub-block to enhance the initial reflection component set of the image sub-block, thereby obtaining the target reflection component set of the image sub-block.

5. The method according to claim 4, characterized in that, The process of determining the adaptive contrast adjustment factor set of the image sub-block, and using the adaptive contrast adjustment factor set of the image sub-block to enhance the initial reflection component set of the image sub-block to obtain the target reflection component set of the image sub-block includes: Determine the global maximum contrast and local contrast set of the image sub-block, and calculate the adaptive contrast adjustment factor set of the image sub-block using the global maximum contrast and local contrast set of the image sub-block. in, Indicates the image sub-block at the target pixel coordinates Adaptive contrast adjustment factor at the location, Indicates the image sub-block at the target pixel coordinates Local contrast at that location This represents the global maximum contrast of the image sub-block. Indicates a very small constant value; The initial reflection component set of the image sub-block is enhanced using the adaptive contrast adjustment factor set of the image sub-block to obtain the enhanced reflection component set of the image sub-block. in, Indicates the image sub-block at the target pixel coordinates Enhanced reflection component at the location, Indicates the image sub-block at the target pixel coordinates The initial reflection component at that location, Indicates the image sub-block at the target pixel coordinates Adaptive contrast adjustment factor at the location; Obtain nonlinear adjustment parameters, and use these parameters to perform nonlinear mapping on the enhanced reflection component set of the image sub-block to obtain the target reflection component set of the image sub-block. in, Indicates the image sub-block at the target pixel coordinates The target reflection component at that location, Indicates the image sub-block at the target pixel coordinates Adaptive contrast adjustment factor at the location, This represents a very small constant value. Indicates the image sub-block at the target pixel coordinates Pixel value at that location, Indicates the image sub-block at the target pixel coordinates The amount of light at that location, This represents the nonlinear adjustment parameter.

6. The method according to claim 1, characterized in that, The step of performing multi-scale wavelet decomposition on each image sub-block to obtain the low-frequency component and weighted high-frequency component of each image sub-block includes: For each image sub-block, a wavelet decomposition operation is performed on the image sub-block to obtain the wavelet decomposition result of the image sub-block. in, This represents the wavelet decomposition result of the r-th image sub-block. This represents the r-th image sub-block. This represents the low-frequency component of the r-th image sub-block. This represents the horizontal high-frequency component of the r-th image sub-block. This represents the high-frequency component in the vertical direction of the r-th image sub-block. This represents the high-frequency component in the diagonal direction of the r-th image sub-block; The high-frequency components in the horizontal, vertical, and diagonal directions of the wavelet decomposition results of the image sub-block are merged to obtain the merged high-frequency components of the image sub-block. in, This represents the merged high-frequency components of the r-th image sub-block. This represents the horizontal high-frequency component of the r-th image sub-block. This represents the high-frequency component in the vertical direction of the r-th image sub-block. This represents the high-frequency component in the diagonal direction of the r-th image sub-block; The detail weighting factor set of the image sub-block is determined using the local brightness mean set and local contrast set of the image sub-block, and the detail weighting factor matrix of the image sub-block is constructed using the detail weighting factor set. in, This represents the target pixel coordinates of the r-th image sub-block. The detail weighting factor, This represents the target pixel coordinates of the r-th image sub-block. The local average brightness at that location, This represents the target pixel coordinates of the r-th image sub-block. Local contrast at that location Indicates a very small constant value; The weighted high-frequency components of the image sub-block are determined by using the detail weighting factor matrix of the image sub-block and merging high-frequency components. in, This represents the weighted high-frequency component of the r-th image sub-block. This represents the detail weighting factor matrix of the r-th image sub-block. This represents the merged high-frequency components of the r-th image sub-block.

7. The method according to claim 1, characterized in that, The nonlinear fusion of the target reflection component set, low-frequency component, and weighted high-frequency component of each image sub-block to obtain an enhanced night view image includes: For each image sub-block, the target reflection component set, low-frequency component, and weighted high-frequency component of the image sub-block are nonlinearly fused to obtain the enhanced image sub-block. in, This represents the target pixel coordinates of the r-th enhanced image sub-block. Pixel value at that location, This represents the target pixel coordinates of the r-th image sub-block. The target reflection component at that location, This represents the weighted high-frequency component of the r-th image sub-block. This represents the low-frequency component of the r-th image sub-block. This represents a very small constant value. Indicates the first fusion index, This indicates the second fusion index. Indicates the exponent of the power transformation; The enhanced image sub-blocks obtained by nonlinearly fusing each of the image sub-blocks are obtained, resulting in multiple enhanced image sub-blocks. The multiple enhanced image sub-blocks are then recombined to obtain the enhanced night view image.

8. The method according to claim 7, characterized in that, The recombining of the plurality of enhanced image sub-blocks to obtain the enhanced night view image includes: The multiple enhanced image sub-blocks are recombined according to the target pixel coordinates of each enhanced image sub-block to obtain the enhanced night view image. in, This refers to the enhanced night view image. This represents the r-th enhanced image sub-block. This indicates the total number of enhanced image sub-blocks.

9. An image enhancement device for local feature overlay and weighted fusion, characterized in that, include: The segmentation module is used to acquire an initial night view image and segment the initial night view image into multiple image sub-blocks; The illumination extraction module is used to perform adaptive edge-protected illumination extraction operations on each of the image sub-blocks to obtain the illumination component set of each of the image sub-blocks; The reflection enhancement module is used to perform initial reflection component separation operation on the illumination component set of each image sub-block using the Retinex model, and to enhance the target reflection component set of each image sub-block by introducing an adaptive contrast adjustment factor set of each image sub-block. The wavelet decomposition module is used to perform multi-scale wavelet decomposition on each of the image sub-blocks to obtain the low-frequency components and weighted high-frequency components of each of the image sub-blocks. The fusion module is used to perform nonlinear fusion of the target reflection component set, low-frequency component and weighted high-frequency component of each image sub-block to obtain an enhanced night view image. The nonlinear fusion includes weighted superposition of the low-frequency component and the target reflection component set, and nonlinear transformation processing of the result of the weighted superposition and the weighted high-frequency component.

10. An apparatus comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 8.

Images