Night infrared monitoring imaging enhancement method and device

Abstract

The invention relates to a night infrared monitoring imaging enhancement method and device, and the method comprises the following steps: obtaining an original infrared image of a night scene through infrared monitoring equipment, carrying out the noise suppression, obtaining a denoised infrared image, and carrying out the contrast enhancement of the image, and generating an enhanced contrast image. Performing target area initial positioning on the original image based on the image to form an initial positioning target frame, and determining an effective target frame through area gray analysis; detail sharpening processing is carried out on the contrast-enhanced image for the frame, a sharpened target image is generated, and then a target contour is extracted to obtain clear target contour information. And finally, performing feature annotation on the contour map to generate a target feature annotation map, and performing background weakening processing on the map to finally obtain an enhanced imaging map which highlights a target and has a soft background, thereby solving the technical problems of fuzzy details and serious noise interference generally existing in the existing infrared monitoring image.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image enhancement, in particular to a night infrared monitoring imaging enhancement method and device. BACKGROUND

[0002] With the rapid development of application fields such as urban security, intelligent transportation and field monitoring, night monitoring as a key link of all-weather monitoring system is becoming increasingly important. The performance of traditional visible light monitoring decreases sharply at night or in low-illumination environment, which is difficult to meet the actual demand. However, the existing infrared monitoring image generally has low contrast, fuzzy details and serious noise interference, especially in complex night scenes, the temperature difference between the target and the background is small, which leads to unclear target edges and difficult-to-identify features, seriously affecting the subsequent identification and analysis. SUMMARY

[0003] The present application aims to at least solve one of the technical problems existing in the prior art. To this end, the present application provides a night infrared monitoring imaging enhancement method, comprising the following steps: An original infrared image of a night scene is collected by an infrared monitoring device, noise suppression processing is performed on the original infrared image to obtain a denoised infrared image, and contrast enhancement processing is performed on the denoised infrared image to obtain an enhanced contrast image; Target positioning is performed based on the enhanced contrast image to obtain a primary positioning target frame, and regional gray scale analysis is performed on the primary positioning target frame to obtain an effective target frame; Based on the effective target frame, detail sharpening processing is performed on the enhanced contrast image to obtain a sharpened target image, and contour extraction processing is performed on the sharpened target image to obtain clear target contour information; Based on the clear target contour information, feature labeling processing is performed to obtain a target feature labeling image, and background weakening processing is performed on the target feature labeling image to obtain an enhanced imaging image.

[0004] Further, the noise suppression processing on the original infrared image to obtain a denoised infrared image comprises: Noise type identification is performed on the original infrared image to obtain a noise distribution heat map, and block division is performed on the noise distribution heat map to obtain a noise block map; Different types of noise blocks in the noise block map are filtered to obtain a preliminary denoised image, and edge fidelity processing is performed on the preliminary denoised image to obtain a denoised infrared image.

[0005] Further, the contrast enhancement processing on the denoised infrared image to obtain an enhanced contrast image comprises: The denoised infrared image is subjected to gray-level histogram statistics to obtain a gray-level distribution histogram, and the gray-level distribution histogram is subjected to dynamic range analysis to obtain a gray-level interval division table. Based on the grayscale interval division table, the dark area interval in the denoised infrared image is stretched to obtain a dark area enhancement image, and the contrast of the mid-gray area interval in the dark area enhancement image is adjusted to obtain a mid-gray enhancement image. Gray-scale suppression is applied to the bright areas in the mid-gray enhancement image to obtain a bright area calibration image. Then, full-image gray-scale fusion is performed on the bright area calibration image to obtain an enhanced contrast image.

[0006] Furthermore, the step of performing target localization based on the enhanced contrast map to obtain an initial target bounding box includes: The enhanced contrast image is compared with the neighboring gray levels to obtain a gray level difference image, and the gray level difference image is binarized to obtain a binary target image. Based on the binary target image, small region culling is performed to obtain a cleaned target image, and morphological erosion is performed on the cleaned target image to obtain an eroded target contour image; The eroded target contour map is matched with the original infrared image by pixel coordinate matching to obtain a matched target map, and the target area is marked on the matched target map to obtain a marked target map; Based on the marked target map, the extreme coordinates of the target region are extracted to obtain the target coordinate group, and the rectangular boundary of the target coordinate group is drawn to obtain the initial positioning target box.

[0007] Furthermore, the step of performing detail sharpening processing on the enhanced contrast image based on the effective target bounding box to obtain a sharpened target image includes: Pixels are extracted from the enhanced contrast map region corresponding to the effective target box to obtain a target region sub-image, and grayscale gradient is calculated on the target region sub-image to obtain a gradient intensity map; The target region sub-image is subjected to edge enhancement processing based on the gradient intensity map to obtain an edge-enhanced sub-image, and the edge-enhanced sub-image is subjected to detail texture extraction to obtain a texture feature map. Based on the texture feature map, the edge enhancement sub-image is subjected to texture fusion processing to obtain a fused sub-image, and the fused sub-image and the enhanced contrast map are then stitched together to obtain a sharpened target image.

[0008] Furthermore, the step of performing feature annotation processing based on the clear target contour information to obtain a target feature annotation map includes: The contour feature points of the clear target contour information are extracted to obtain the coordinates of the feature points, and the geometric parameters of the feature point coordinates are calculated to obtain the target geometric parameter table; Based on the target geometric parameter table, feature regions in the clear target contour information are divided to obtain feature sub-region maps, and gray-scale feature statistics are performed on the feature sub-region maps to obtain sub-region gray-scale tables. Based on the grayscale table of the sub-region, key features in the feature sub-region map are identified, and feature attributes are labeled on the key features to obtain a feature attribute table; Based on the feature attribute table, key features are visualized and overlaid to obtain an overlaid feature map. The overlaid feature map is then integrated with annotation information to obtain a target feature annotation map.

[0009] Furthermore, the step of performing background weakening processing on the target feature annotation map to obtain an enhanced imaging map includes: A mask is generated for the target region in the target feature annotation map to obtain a target mask map, and the target mask map and the target feature annotation map are superimposed pixel by pixel to obtain a target preserved map; The background area in the target retained image is grayscale suppressed to obtain a background suppressed image, and the background suppressed image is Gaussian blurred to obtain a blurred background image. Image fusion is performed based on the target preserved image and the blurred background image to obtain an initial enhanced image, and edge transition optimization is performed on the initial enhanced image to obtain an enhanced imaging image.

[0010] The present invention also provides a nighttime infrared surveillance imaging enhancement device, comprising: The acquisition module is used to acquire raw infrared images of nighttime scenes through infrared monitoring equipment, perform noise suppression processing on the raw infrared images to obtain a denoised infrared image, and perform contrast enhancement processing on the denoised infrared image to obtain an enhanced contrast image. The positioning module is used to locate the target based on the enhanced contrast map to obtain an initial positioning target box, and to perform regional grayscale analysis on the initial positioning target box to obtain an effective target box; The sharpening module is used to perform detail sharpening processing on the enhanced contrast image based on the effective target box to obtain a sharpened target image, and to perform contour extraction processing on the sharpened target image to obtain clear target contour information; The annotation module is used to perform feature annotation processing based on the clear target contour information to obtain a target feature annotation map, and to perform background weakening processing on the target feature annotation map to obtain an enhanced imaging map.

[0011] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the methods described above.

[0012] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of any of the methods described above.

[0013] This invention provides a nighttime infrared surveillance imaging enhancement method, comprising the following steps: acquiring raw infrared images of a nighttime scene using an infrared surveillance device; performing noise suppression processing on the raw infrared images to obtain a denoised infrared image; and performing contrast enhancement processing on the denoised infrared image to obtain an enhanced contrast image; performing target localization based on the enhanced contrast image to obtain an initial target bounding box; and performing regional grayscale analysis on the initial target bounding box to obtain an effective target bounding box; performing detail sharpening processing on the enhanced contrast image based on the effective target bounding box to obtain a sharpened target image; and performing contour extraction processing on the sharpened target image to obtain clear target contour information; based on... The clear target contour information is processed by feature annotation to obtain a target feature annotation map. The target feature annotation map is then subjected to background weakening processing to obtain an enhanced imaging map. This solves the technical problems of blurred details and severe noise interference in existing infrared surveillance images. It enables the initial positioning of the original infrared image by utilizing the high discriminative characteristics of the enhanced contrast map, avoiding the deviation caused by direct positioning on the low-contrast original image. Furthermore, by performing regional grayscale analysis on the initially located target box, false targets or noise interference areas are eliminated, and effective target boxes with grayscale features that better match the characteristics of the real target are selected. This achieves a more accurate and robust target area recognition effect. Attached Figure Description

[0014] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a flowchart illustrating the steps of the nighttime infrared surveillance imaging enhancement method in an embodiment of the present invention. Figure 2 This is a structural block diagram of the nighttime infrared monitoring imaging enhancement device in an embodiment of the present invention; Figure 3 This is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0015] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0016] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0017] In the description of this invention, it should be understood that the orientation descriptions, such as up, down, front, back, left, right, etc., are based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention 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 limiting this invention.

[0018] In the description of this invention, "several" means one or more, "more than" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. The use of "first" and "second" in the description is merely for distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.

[0019] In the description of this invention, unless otherwise explicitly defined, terms such as "setting," "installing," and "connecting" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this invention in conjunction with the specific content of the technical solution.

[0020] The embodiments of this application will be further described below with reference to the accompanying drawings.

[0021] Reference Figure 1 This invention provides a nighttime infrared surveillance imaging enhancement method, comprising the following steps: Step S1: Collect original infrared images of the night scene using infrared monitoring equipment, perform noise suppression processing on the original infrared images to obtain a denoised infrared image, and perform contrast enhancement processing on the denoised infrared image to obtain an enhanced contrast image.

[0022] Specifically, raw infrared images of nighttime scenes are acquired through infrared monitoring equipment. These raw images, often under conditions of low light, small temperature differences, or equipment sensitivity limitations, typically contain random pixel fluctuations caused by sensor thermal noise and environmental interference. Therefore, noise suppression processing is necessary to obtain a more stable denoised infrared image. In practice, nonlocal mean filtering or wavelet thresholding algorithms can be used to process the raw infrared image pixel by pixel. By calculating the similarity weights of pixels within local neighborhoods, isolated noise points are suppressed while preserving edges and texture structures as much as possible. For example, in urban nighttime road monitoring scenes, pedestrian or vehicle edges often appear as ghosting or broken due to noise. After nonlocal mean filtering, the pixel values ​​in these areas are reasonably smoothed, resulting in a clear and stable denoised infrared image. After obtaining the denoised infrared image, due to the inherent narrow grayscale dynamic range and insignificant temperature difference between the target and background in infrared imaging, the overall image contrast is low. Therefore, further noise suppression processing is needed. Further contrast enhancement processing is performed. This processing can be achieved through adaptive histogram equalization (CLAHE), which divides the grayscale histogram of the denoised infrared image into multiple sub-blocks and performs histogram equalization independently within each sub-block to avoid local over-enhancement caused by global equalization. At the same time, a contrast limit threshold is set to prevent noise from being excessively amplified. For example, in the same urban nighttime road monitoring scene, after CLAHE processing, the difference between the road surface and parked vehicles, which were originally similar in grayscale, is effectively amplified, and the body contours of pedestrians show more obvious brightness changes in the image, ultimately generating an enhanced contrast map with higher visual discrimination. The entire process starts from the aforementioned original infrared image, and successively goes through noise suppression processing and contrast enhancement processing. The output enhanced contrast map not only retains the true target structure information, but also significantly improves the image quality that can be used for subsequent analysis, providing a reliable data foundation for target area localization based on this map.

[0023] Step S2: Based on the enhanced contrast map, target localization is performed to obtain an initial target bounding box, and regional grayscale analysis is performed on the initial target bounding box to obtain an effective target bounding box.

[0024] Specifically, region localization is performed on the original infrared image based on the enhanced contrast map. Since the grayscale difference between the target and the background is more obvious after contrast enhancement processing, the image can be used as a guide to improve localization accuracy. In practice, firstly, a threshold-based segmentation method is used on the enhanced contrast map to determine potential target regions. By setting an adaptive threshold, pixel regions above the threshold are marked as candidate target regions. Then, morphological closing operations are performed on these regions to fill internal holes and connect adjacent breakpoints. Next, connected component analysis is used to extract the boundary boxes of each independent region, thus obtaining the initial localization target boxes. These initial localization target boxes contain possible target location information, but since artifacts or background interference regions may be introduced during the enhancement process, their effectiveness needs to be further verified. Next, region grayscale analysis is performed on the initial localization target boxes. This analysis process needs to be traced back to the original infrared image because the original infrared image retains the true temperature distribution. The feature distribution avoids distortion caused by the enhancement process. Specifically, the region corresponding to each initial target box is extracted from the original infrared image, and the gray-scale mean, variance, and gradient change density of its internal pixels are calculated. If the gray-scale mean of a region is significantly higher than the surrounding background and the variance is within a reasonable range (neither too uniform nor too chaotic), and the gradient density reaches a preset threshold, then the region is determined to have real heat source characteristics and is retained as a valid target box. For example, in a nighttime urban road monitoring scene, a parked car appears as a bright rectangular area in the enhanced contrast image, forming an initial target box. However, this box may contain part of the road surface reflecting heat radiation area. By performing gray-scale analysis on the corresponding region of the box in the original infrared image, it is found that the gray-scale value of the left part is close to the ambient temperature and the gradient change is weak. Therefore, this part is removed, and only the right part of the car body with stable high-temperature characteristics is retained as the final valid target box, thus ensuring that subsequent processing is based on the accurate target location.

[0025] Step S3: Based on the effective target bounding box, perform detail sharpening processing on the enhanced contrast image to obtain a sharpened target image, and perform contour extraction processing on the sharpened target image to obtain clear target contour information.

[0026] Specifically, to perform detail sharpening on the enhanced contrast image based on the effective target boxes, it is first necessary to determine the specific location of the effective target boxes in the enhanced contrast image. Then, local detail sharpening operations are performed on these areas to highlight the target edges and texture features. Specifically, after determining the effective target boxes, methods such as the Laplacian operator or nonlinear sharpening filters can be used to process the pixels within each effective target box. By calculating the difference between each pixel in the image and its surrounding pixels, the brightness value at the edges is increased, thereby achieving a sharpening effect. For example, in urban nighttime road monitoring scenarios, for the vehicle outline that has been identified as an effective target box, applying the Laplacian operator can enhance key details such as the edges of the headlights, body lines, and license plate numbers, making the originally blurry features more clearly visible, thus obtaining a sharpened target image. Next, contour extraction is performed on the sharpened target image to accurately depict the boundary of the target object from the sharpened image. This process can be achieved using the Canny edge detection algorithm. This algorithm first smooths the image using a Gaussian filter to reduce noise interference, then calculates the intensity gradient and direction of the image, applies non-maximum suppression technology to refine the edges, and finally determines the final edge using a double threshold algorithm. In the above application scenario, for the sharpened vehicle image, the Canny algorithm can effectively draw continuous and accurate contour lines along the outer edge of the vehicle, the window frame, and even finer details such as the rearview mirror, thus forming clear target contour information. This not only enhances the visual separation between the target and the background but also provides high-quality basic data for subsequent target feature annotation and background weakening, ensuring that the final enhanced image can more accurately reflect the real situation in the monitoring scene. Throughout the process, from using effective target boxes to guide detail sharpening to obtaining clear target contour information through contour extraction technology, each step is closely linked, working together to improve the readability and usability of infrared monitoring images.

[0027] Step S4: Based on the clear target contour information, perform feature annotation processing to obtain a target feature annotation map, and perform background weakening processing on the target feature annotation map to obtain an enhanced imaging map.

[0028] Specifically, feature annotation processing based on the clear target contour information first requires identifying and determining the key feature regions in the image that need to be annotated. For example, in urban nighttime road monitoring scenarios, for the clear target contour information of vehicles obtained after sharpening and contour extraction, key parts such as license plates, headlights, and unique markings on the vehicle body can be further located. Next, these key parts are annotated using different colors or marking symbols to ensure that each annotated feature is visually prominent. For example, a red box is used to mark the license plate position, a yellow circle is used to mark the headlight position, and specific symbols are added to scratches or special markings on the vehicle body to form a target feature annotation map. This process not only helps improve the accuracy of image analysis but also facilitates subsequent automated recognition. Then, the target feature annotation map is subjected to background weakening processing to reduce the interference of background information on the target object and enhance the prominence of the target object in the image. This can be achieved by adjusting parameters such as color saturation, contrast, and brightness. For example, in the above application scenario, the brightness and contrast of all parts of the image except the annotated target can be reduced, making the background appear somewhat blurred while maintaining high brightness and high contrast for the vehicle and the annotated features. Furthermore, edge detection technology can be used to further enhance the clarity of the boundary between the target and the background, making the target object appear more clearly to the observer. For example, by increasing the width and depth of the vehicle's outline, it can stand out against a dark background. Through a series of fine adjustments, an enhanced image is finally obtained that retains important features while effectively weakening background information. This image not only clearly displays the vehicle and its key features but also greatly avoids interference from cluttered background information, improving the overall readability and usability of the image, making it particularly suitable for the rapid and accurate identification and tracking of targets in the field of security monitoring. Throughout the process, from feature annotation to background weakening, each step is closely linked and works together to improve the clarity and effectiveness of infrared surveillance images.

[0029] In a specific embodiment, the step of performing noise suppression processing on the original infrared image to obtain a denoised infrared image includes: The original infrared image is subjected to noise type identification to obtain a noise distribution heatmap, and the noise distribution heatmap is divided into blocks to obtain a noise block map; Different types of noise blocks in the noise block image are filtered in a targeted manner to obtain a preliminary denoised image. The preliminary denoised image is then subjected to edge fidelity processing to obtain a denoised infrared image.

[0030] Specifically, noise suppression processing of the original infrared image first requires noise type identification to obtain a noise distribution heatmap. This process involves analyzing the grayscale value changes of each pixel in the image and considering the different characteristics of different types of noise. For example, Gaussian noise typically manifests as small, randomly distributed grayscale fluctuations throughout the image, while salt-and-pepper noise appears as alternating black and white bright spots or dark spots. These characteristics effectively distinguish noise from background information. Next, to more accurately locate and process noise, the noise distribution heatmap needs to be divided into blocks to obtain a noise block map. This step involves dividing the entire image into several sub-regions, each marked according to the intensity, density, and distribution pattern of its internal noise, thus forming a map reflecting the noise distribution. For instance, in a security monitoring system, for infrared video streams captured at night, the above method can be used to identify and divide noise areas caused by changes in ambient light. Next, targeted filtering is performed on different types of noise blocks in the noise block image to obtain a preliminary denoised image. At this stage, for blocks marked as Gaussian noise, mean filtering or Gaussian filtering can be used to smooth the pixel values ​​in these areas. This is achieved by calculating the average or weighted average of the pixels surrounding a given pixel to replace that pixel value, thereby reducing the noise impact. For blocks marked as salt-and-pepper noise, mean filtering is more suitable. This involves taking the median of all pixel values ​​within a given point and its neighborhood as the new value for that point, effectively removing isolated bright and dark spots without damaging image edge information. Furthermore, when encountering mixed noise, it may be necessary to combine multiple filtering algorithms. For example, bilateral filtering can be used first to reduce noise while preserving edge information, followed by adaptive filtering to dynamically adjust filtering parameters based on local noise levels to ensure optimal denoising results. For instance, in industrial inspection scenarios, this combined strategy can significantly improve defect recognition rates when dealing with infrared images containing complex noise. Finally, to compensate for potential edge blurring during the filtering process, edge fidelity processing is required on the initial denoised image to obtain the final denoised infrared image. This involves using edge detection algorithms, such as the Sobel operator or the Canny algorithm, to first locate the boundaries of objects in the image, and then enhance the contrast at these boundaries to make the object outlines clearer and sharper. Simultaneously, advanced techniques such as non-local means filtering can be introduced. This not only considers the spatial proximity of pixels but also the similarity between pixels, preserving details well even far from the target edge and avoiding over-smoothing. For example, in the field of medical imaging, edge fidelity processing of the patient's infrared thermal image after the above steps allows doctors to more accurately determine the boundary between lesions and normal tissue, improving diagnostic accuracy.Throughout the entire process, from noise type identification to block segmentation, targeted filtering, and edge fidelity processing, each step is closely linked and works together to improve the quality of infrared images, enabling them to play a greater role in various application scenarios.

[0031] In a specific embodiment, the step of performing contrast enhancement processing on the denoised infrared image to obtain an enhanced contrast image includes: The denoised infrared image is subjected to gray-level histogram statistics to obtain a gray-level distribution histogram, and the gray-level distribution histogram is subjected to dynamic range analysis to obtain a gray-level interval division table. Based on the grayscale interval division table, the dark area interval in the denoised infrared image is stretched to obtain a dark area enhancement image, and the contrast of the mid-gray area interval in the dark area enhancement image is adjusted to obtain a mid-gray enhancement image. Gray-scale suppression is applied to the bright areas in the mid-gray enhancement image to obtain a bright area calibration image. Then, full-image gray-scale fusion is performed on the bright area calibration image to obtain an enhanced contrast image.

[0032] Specifically, to enhance the contrast of the denoised infrared image, the first step is to perform grayscale histogram statistics. This involves calculating the number of pixels at each grayscale level to obtain the grayscale distribution histogram. This process reveals the brightness distribution characteristics of the image. For example, in urban nighttime road monitoring scenarios, since the background is mostly dark while target objects such as vehicles or pedestrians are relatively bright, the histogram often exhibits a high-to-low shape, meaning a large number of pixels are concentrated in the darker and brighter grayscale levels, while there are fewer pixels in the medium grayscale levels. Based on this grayscale distribution histogram, dynamic range analysis is further performed to identify the effective grayscale intervals in the image and divide them into different grayscale intervals, forming a grayscale interval division table. This step helps to clarify which areas need to be enhanced. For example, in the above scenario, 0-64 can be considered the dark area interval, 65-192 the medium gray area interval, and 193-255 the bright area interval. Next, based on the grayscale interval division table, grayscale stretching is performed on the dark areas of the denoised infrared image to obtain a dark enhancement image. Specifically, linear or nonlinear transformations (such as Gamma correction) are applied to the pixel values ​​within the dark areas to appropriately amplify the originally low grayscale values, thereby improving the visibility of dark details. For example, in nighttime surveillance footage, after grayscale stretching, the cracks or small objects hidden in the shadows on the road become clearer, enhancing the overall image information. Subsequently, contrast adjustment is performed on the mid-gray areas of the dark enhancement image. By adjusting the degree of difference in pixel values ​​within this area, more detailed features are highlighted, generating a mid-gray enhancement image. Here, histogram equalization can be used, which not only increases the dynamic range of the mid-grayscale level but also makes the transition between different grayscale values ​​smoother and more natural. For example, in urban surveillance videos, the different textures of building surface materials can be more clearly displayed. Then, grayscale suppression is applied to the bright areas in the mid-gray enhancement image to avoid overexposure and improve visual comfort, thus obtaining a bright area calibration image. For areas with high brightness, their grayscale values ​​are appropriately reduced, which can prevent the loss of key information due to overexposure and reduce the stimulation to the observer's eyes. In practical application scenarios, such as when facing strong light spots caused by streetlights, reasonable suppression of the bright areas can preserve the presence of the light without making the whole picture glaring, while also allowing for a clearer view of the area around the light source.Finally, the brightness calibration map undergoes full-image grayscale fusion, which integrates the adjustment results of the previously processed dark, mid-gray, and bright areas. This ensures that the entire image, from the darkest to the brightest parts, receives appropriate enhancement, ultimately forming an enhanced contrast map. This process may require weighted averaging or other fusion strategies to ensure smooth and consistent transitions between different areas. For example, in nighttime surveillance scenarios, such full-image grayscale fusion not only clearly displays pedestrians, vehicles, and their surrounding environmental details but also ensures the consistency and coherence of the overall image, greatly improving the efficiency and reliability of the surveillance system. Throughout the entire process, from grayscale histogram statistics to the final full-image grayscale fusion, each step is closely linked, working together to improve the contrast and readability of the infrared image.

[0033] In a specific embodiment, the step of locating the target based on the enhanced contrast map to obtain an initial target bounding box includes: The enhanced contrast image is compared with the neighboring gray levels to obtain a gray level difference image, and the gray level difference image is binarized to obtain a binary target image. Based on the binary target image, small region culling is performed to obtain a cleaned target image, and morphological erosion is performed on the cleaned target image to obtain an eroded target contour image; The eroded target contour map is matched with the original infrared image by pixel coordinate matching to obtain a matched target map, and the target area is marked on the matched target map to obtain a marked target map; Based on the marked target map, the extreme coordinates of the target region are extracted to obtain the target coordinate group, and the rectangular boundary of the target coordinate group is drawn to obtain the initial positioning target box.

[0034] Specifically, to perform region localization on the original infrared image based on the enhanced contrast map, the enhanced contrast map first needs to be compared with neighboring grayscale values ​​to identify the boundaries and changes between different regions, thus obtaining a grayscale difference map. In this process, for each pixel, the difference in grayscale value between it and pixels within a certain range around it is calculated, and these differences are used as new image pixel values ​​to construct a grayscale difference map reflecting local brightness changes. For example, when monitoring a parking lot at night, the boundary between the vehicle and the background will show a significant grayscale difference due to their different infrared reflection capabilities, providing a basis for subsequent steps. Next, the grayscale difference map is binarized to simplify image information and highlight potential target areas, thereby generating a binary target map. Here, a threshold needs to be selected; all pixels below this threshold are set to 0 (or background), and pixels above this threshold are set to 1 (or foreground), thus distinguishing target and non-target areas. In the parking lot example above, an appropriate threshold can be selected by analyzing the histogram of the grayscale difference map, allowing the vehicle outline to be clearly separated from the background, forming a clear black-and-white binary target map. Then, based on the binary target image, small region culling is performed to remove areas that may be composed of noise or other irrelevant small spots, thus obtaining a clutter-free target image. This is because some isolated small areas are likely caused by noise during image processing rather than the actual target object. Subsequently, morphological erosion is performed on the clutter-free target image to eliminate small connecting parts by shrinking the boundaries of foreground objects, ultimately obtaining an eroded target contour image. In practice, such as when processing nighttime parking lot images, there may be some false edges or tiny noise points that interfere with the target detection results. In this case, erosion can effectively purify the target contour, ensuring that only the real large targets (such as cars) are retained. Next, the eroded target contour image is matched with the original infrared image using pixel coordinates to recover the specific position of the target object in the original image, thereby generating a matched target image. This process involves remapping the target contour after a series of processing steps back to the original image space for subsequent analysis and understanding. Continuing with the aforementioned parking lot example, coordinate matching can accurately determine the specific parking position of each car in the parking lot, preparing for further operations.Finally, the matched target map is marked with target regions. A marked target map is generated by assigning a unique identifier to each individual target. Based on this, the extreme coordinates of the target regions are extracted from the marked target maps, i.e., the extreme positions of each target in the top, bottom, left, and right directions are found, thus obtaining a target coordinate set. A rectangular boundary is then drawn for the target coordinate set, ultimately obtaining the initial target bounding box. For example, in a parking lot monitoring application scenario, once the location of each vehicle is determined, a bounding box can be drawn for it. This not only clearly marks the presence of the vehicle but also provides key information about the vehicle's size and location, which is of great significance for subsequent advanced applications such as vehicle counting and behavior analysis. Throughout the entire process, from the generation of the grayscale difference map to the determination of the initial target bounding box, each step is closely linked, working together to achieve effective positioning and recognition of targets in the original infrared image.

[0035] In a specific embodiment, the step of performing detail sharpening processing on the enhanced contrast image based on the effective target box to obtain a sharpened target image includes: Pixels are extracted from the enhanced contrast map region corresponding to the effective target box to obtain a target region sub-image, and grayscale gradient is calculated on the target region sub-image to obtain a gradient intensity map; The target region sub-image is subjected to edge enhancement processing based on the gradient intensity map to obtain an edge-enhanced sub-image, and the edge-enhanced sub-image is subjected to detail texture extraction to obtain a texture feature map. Based on the texture feature map, the edge enhancement sub-image is subjected to texture fusion processing to obtain a fused sub-image, and the fused sub-image and the enhanced contrast map are then stitched together to obtain a sharpened target image.

[0036] Specifically, based on the effective target bounding box, the enhanced contrast image is sharpened by extracting pixels from the enhanced contrast image region corresponding to the effective target bounding box to obtain a more refined target region sub-image. Then, the grayscale gradient of the target region sub-image is calculated to obtain a gradient intensity map that reflects the degree of brightness change in the image. In this process, for each pixel, the gradient direction and magnitude of the pixel are determined by calculating the grayscale difference between it and the surrounding pixels. For example, when monitoring a parking lot at night, the boundary between the vehicle outline and the background will show obvious grayscale gradient changes due to the different abilities of reflecting infrared light, which provides key information for subsequent steps.

[0037] Next, edge enhancement processing is performed on the target region sub-image based on the gradient intensity map to highlight the boundary features of the target object, thereby generating an edge-enhanced sub-image. The method used here can be to enhance the high-value regions (i.e., edges) in the gradient intensity map while keeping the non-edge parts unchanged. In the parking lot example above, the edges of the vehicles can be more clearly separated from the background in this way, forming a sharp contrast. Subsequently, detailed texture extraction is performed on the edge-enhanced sub-image to capture subtle structural changes in the image, thereby obtaining a texture feature map. This process involves analyzing the gray-level distribution characteristics within a small range in the image to identify local features that represent specific texture patterns, such as the texture of the parking lot floor or the details of the car surface. Then, based on the texture feature map, the edge enhancement sub-image is subjected to texture fusion processing, that is, the edge information in the original edge enhancement sub-image is combined with the newly extracted texture features to generate richer image content, and thus obtain the fused sub-image. This step requires careful balancing of the ratio of edge information and texture features to ensure that the two can coexist harmoniously rather than overlap each other. For example, in the application scenario of parking lots, it is necessary not only to emphasize the outline of the vehicle, but also to retain detailed features such as license plate number and body decoration, so that the final image is both clear and contains sufficient detailed information. Finally, the fused sub-image and the enhanced contrast image are stitched together to restore the overall consistency of the image, thereby generating the sharpened target image. During this process, attention must be paid to the consistency of color and brightness between the fused sub-image and the original image to avoid unnatural transitions. Continuing with the parking lot example, after sharpening a single vehicle image, it is re-embedded into the panoramic image of the entire parking lot to ensure that each vehicle is displayed in its best condition and seamlessly integrated with its surroundings. Through this process, from pixel extraction to background stitching, each step is closely linked, working together to effectively sharpen the target in the original enhanced contrast image. This not only improves the visual quality of the image but also lays a solid foundation for further image analysis and understanding. Throughout this process, attention to detail and meticulous operation significantly improve the accuracy and reliability of target detection and recognition, which is of great significance for many practical applications. For example, in intelligent transportation systems, improving the recognition accuracy of vehicles in parking lots can greatly improve parking management efficiency and service experience.

[0038] In a specific embodiment, the step of performing feature annotation processing based on the clear target contour information to obtain a target feature annotation map includes: The contour feature points of the clear target contour information are extracted to obtain the coordinates of the feature points, and the geometric parameters of the feature point coordinates are calculated to obtain the target geometric parameter table; Based on the target geometric parameter table, feature regions in the clear target contour information are divided to obtain feature sub-region maps, and gray-scale feature statistics are performed on the feature sub-region maps to obtain sub-region gray-scale tables. Based on the grayscale table of the sub-region, key features in the feature sub-region map are identified, and feature attributes are labeled on the key features to obtain a feature attribute table; Based on the feature attribute table, key features are visualized and overlaid to obtain an overlaid feature map. The overlaid feature map is then integrated with annotation information to obtain a target feature annotation map.

[0039] Specifically, feature annotation processing based on the clear target contour information first requires extracting contour feature points. This is done by identifying the edges and boundaries of objects in the image to determine their coordinates, thereby obtaining a series of key location points representing the shape of the object. These feature points can be corners, inflection points on curves, or any location with significant changes. For example, when analyzing a photo of a car, algorithms can automatically locate key parts such as wheels and window frames, thereby obtaining a set of coordinate points that accurately describe the car's shape. Geometric parameters, including length, angle, and area, are then calculated for these feature point coordinates to form a comprehensive target geometric parameter table reflecting the object's morphological characteristics. This step not only helps to accurately quantify the object's size but also provides a basis for subsequent region division. Next, based on the target geometric parameter table, the feature regions in the clear target contour information are divided. According to the calculated geometric parameters, the image is segmented into multiple sub-regions with similar characteristics, thereby generating a feature sub-region map. In the above example of a car, different feature sub-regions can be defined according to different parts of the vehicle, such as the body, windows, and wheels. Then, grayscale feature statistics are performed on each feature sub-region map to record the distribution of grayscale values ​​of pixels in each sub-region, thereby constructing a sub-region grayscale table. This step helps to gain a deeper understanding of the brightness differences between different parts, such as the obvious contrast that may exist between the windows and the body, providing data support for identifying key features.

[0040] Subsequently, based on the grayscale table of the sub-regions, key features in the feature sub-region map are identified. By analyzing the grayscale information of each sub-region, it is determined which regions contain details crucial for object identification, and these are marked to form key features. Continuing with the example of a car, license plate numbers, brand logos, etc., are important key features, often possessing unique grayscale patterns or textures that are easily distinguishable from the background. Next, feature attributes are labeled for these key features, recording the specific attributes of each key feature in detail, such as color, size, and shape, thereby creating a detailed feature attribute table. This process ensures that even against complex backgrounds, each important element can be accurately identified and described. Finally, based on the feature attribute table, feature visualization overlay is performed on the key features, visually displaying all the marked key features and their attributes on the original image to generate an overlaid feature map. The purpose of this is to allow users to see the analysis results at a glance, such as quickly locating a specific vehicle in a monitoring system. Furthermore, the overlaid feature map is annotated and integrated, concentrating scattered information to form a unified target feature annotation map. In this process, it is necessary not only to ensure the completeness and accuracy of the information but also to consider how to effectively present it to the user for easy understanding and use. The entire process is tightly integrated, from feature point extraction to the final integration of labeled information. Each step is dedicated to improving the accuracy and efficiency of image analysis, enabling detailed and accurate analytical results even in highly complex scenes through meticulous operation. For example, in intelligent traffic management systems, this technology can be used to monitor road conditions in real time and track specific vehicles, significantly improving management efficiency and service quality.

[0041] In a specific embodiment, the step of performing background weakening processing on the target feature annotation map to obtain an enhanced imaging map includes: A mask is generated for the target region in the target feature annotation map to obtain a target mask map, and the target mask map and the target feature annotation map are superimposed pixel by pixel to obtain a target preserved map; The background area in the target retained image is grayscale suppressed to obtain a background suppressed image, and the background suppressed image is Gaussian blurred to obtain a blurred background image. Image fusion is performed based on the target preserved image and the blurred background image to obtain an initial enhanced image, and edge transition optimization is performed on the initial enhanced image to obtain an enhanced imaging image.

[0042] Specifically, the process of background weakening of the target feature annotation map to obtain an enhanced image first involves generating a mask for the target region in the target feature annotation map. This process is achieved by identifying the contours of the target of interest in the image and creating a transparent or semi-transparent mask that matches the shape of the target, i.e., a target mask map, based on these contour information. Then, the target mask map is pixel-wise superimposed with the original target feature annotation map. This step aims to preserve the original details and color information of the target region while preparing for subsequent background processing, thereby generating a target-preserved image. For example, when analyzing an image containing a car, the specific location and boundaries of the vehicle can be determined first, then a mask covering only the car portion can be generated and superimposed on the original image to ensure that the information of the car portion is completely preserved. Next, grayscale suppression is applied to the background area in the target image. This step aims to reduce the visual importance of background elements relative to the target object, allowing the viewer's attention to focus more on the target. Specifically, the grayscale values ​​of the background pixels are adjusted to make them darker or lighter, thereby weakening the background and creating a suppressed background image. To further blur unnecessary details in the background, Gaussian blur processing is applied to the suppressed background image. By applying a Gaussian filter, the transitions between background pixels are smoothed, noise is reduced, and edges are softened, ultimately generating a blurred background image. For example, in the car image example above, grayscale suppression can make background elements such as roads and trees less conspicuous than the car itself, and then Gaussian blur technology can be used to soften the background, avoiding distraction due to an overly cluttered background. Subsequently, image fusion is performed based on the target-preserved image and the blurred background image. This is a crucial step, requiring the precise combination of the sharp target-preserved image and the processed blurred background image to produce an initial enhanced image that highlights the target while maintaining a harmonious background. During this process, careful consideration must be given to the balance of color, brightness, and contrast between the two images to ensure the fused image looks natural and undistorted. Finally, edge transition optimization is performed on the initial enhanced image. This step aims to improve the edge transition effect between the target and the background, making the boundary between them more natural and smooth, eliminating any potentially harsh edges or artifacts, thereby obtaining the final enhanced image. For example, in the application scenario of car images, after fusion, some unnatural lines may be found at the junction of the car and the background. In this case, edge transition optimization technology can make these lines softer, improving the overall visual effect. The entire process starts with mask generation, going through pixel stacking, grayscale suppression, Gaussian blurring, and finally image fusion and edge transition optimization. Each step is closely linked, working together to improve the prominence of the target object in the image while ensuring that the background does not excessively interfere with the viewer's vision. This technology is particularly suitable for situations where it is necessary to emphasize a specific object while downplaying its surroundings, such as in advertising design, product displays, or intelligent monitoring systems that focus attention on a particular individual or object.Through meticulous operation and precise parameter adjustment, the target can be effectively highlighted even against complex backgrounds, providing clear, aesthetically pleasing, and easy-to-interpret visual results.

[0043] The above describes the nighttime infrared surveillance imaging enhancement method in the embodiments of the present invention. The following describes the nighttime infrared surveillance imaging enhancement device in the embodiments of the present invention. Please refer to [link / reference]. Figure 2 One embodiment of the nighttime infrared surveillance imaging enhancement device of the present invention includes: The acquisition module 21 is used to acquire original infrared images of nighttime scenes through infrared monitoring equipment, perform noise suppression processing on the original infrared images to obtain a denoised infrared image, and perform contrast enhancement processing on the denoised infrared image to obtain an enhanced contrast image. The positioning module 22 is used to locate the target based on the enhanced contrast map to obtain an initial positioning target box, and to perform regional grayscale analysis on the initial positioning target box to obtain an effective target box; The sharpening module 23 is used to perform detail sharpening processing on the enhanced contrast image based on the effective target box to obtain a sharpened target image, and to perform contour extraction processing on the sharpened target image to obtain clear target contour information. The annotation module 24 is used to perform feature annotation processing based on the clear target contour information to obtain a target feature annotation map, and to perform background weakening processing on the target feature annotation map to obtain an enhanced imaging map.

[0044] In this embodiment, the specific implementation of each unit in the above device embodiment is described in the above method embodiment, and will not be repeated here.

[0045] like Figure 3 As shown in the diagram, this embodiment of the invention provides a structural schematic block diagram of a computer device, including: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the above-described nighttime infrared surveillance imaging enhancement method.

[0046] It is evident that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented in this device embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0047] Furthermore, this application also discloses a computer program product or computer program stored in a computer-readable storage medium. A processor of a computer device can read the computer program from the computer-readable storage medium and execute the computer program, causing the computer device to perform the aforementioned nighttime infrared surveillance imaging enhancement method. Similarly, the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0048] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A method for enhancing nighttime infrared surveillance imaging, characterized in that, Includes the following steps: The original infrared image of the night scene is acquired by infrared monitoring equipment. The original infrared image is subjected to noise suppression processing to obtain a denoised infrared image. The denoised infrared image is then subjected to contrast enhancement processing to obtain an enhanced contrast image. Target localization is performed based on the enhanced contrast map to obtain an initial target bounding box, and regional grayscale analysis is performed on the initial target bounding box to obtain an effective target bounding box; Based on the effective target bounding box, the enhanced contrast image is subjected to detail sharpening processing to obtain a sharpened target image, and the sharpened target image is subjected to contour extraction processing to obtain clear target contour information. Based on the clear target contour information, feature annotation processing is performed to obtain a target feature annotation map, and the target feature annotation map is then subjected to background weakening processing to obtain an enhanced imaging map.

2. The nighttime infrared surveillance imaging enhancement method according to claim 1, characterized in that, The step of performing noise suppression processing on the original infrared image to obtain a denoised infrared image includes: The original infrared image is subjected to noise type identification to obtain a noise distribution heatmap, and the noise distribution heatmap is divided into blocks to obtain a noise block map; Different types of noise blocks in the noise block image are filtered in a targeted manner to obtain a preliminary denoised image. The preliminary denoised image is then subjected to edge fidelity processing to obtain a denoised infrared image.

3. The nighttime infrared surveillance imaging enhancement method according to claim 1, characterized in that, The step of performing contrast enhancement processing on the denoised infrared image to obtain an enhanced contrast image includes: The denoised infrared image is subjected to gray-level histogram statistics to obtain a gray-level distribution histogram, and the gray-level distribution histogram is subjected to dynamic range analysis to obtain a gray-level interval division table. Based on the grayscale interval division table, the dark area interval in the denoised infrared image is stretched to obtain a dark area enhancement image, and the contrast of the mid-gray area interval in the dark area enhancement image is adjusted to obtain a mid-gray enhancement image. Gray-scale suppression is applied to the bright areas in the mid-gray enhancement image to obtain a bright area calibration image. Then, full-image gray-scale fusion is performed on the bright area calibration image to obtain an enhanced contrast image.

4. The nighttime infrared surveillance imaging enhancement method according to claim 1, characterized in that, The step of locating the target based on the enhanced contrast map to obtain the initial target bounding box includes: The enhanced contrast image is compared with the neighboring gray levels to obtain a gray level difference image, and the gray level difference image is binarized to obtain a binary target image. Based on the binary target image, small region culling is performed to obtain a cleaned target image, and morphological erosion is performed on the cleaned target image to obtain an eroded target contour image; The eroded target contour map is matched with the original infrared image by pixel coordinate matching to obtain a matched target map, and the target area is marked on the matched target map to obtain a marked target map; Based on the marked target map, the extreme coordinates of the target region are extracted to obtain the target coordinate group, and the rectangular boundary of the target coordinate group is drawn to obtain the initial positioning target box.

5. The nighttime infrared surveillance imaging enhancement method according to claim 1, characterized in that, The step of performing detail sharpening processing on the enhanced contrast image based on the effective target bounding box to obtain a sharpened target image includes: Pixels are extracted from the enhanced contrast map region corresponding to the effective target box to obtain a target region sub-image, and grayscale gradient is calculated on the target region sub-image to obtain a gradient intensity map; The target region sub-image is subjected to edge enhancement processing based on the gradient intensity map to obtain an edge-enhanced sub-image, and the edge-enhanced sub-image is subjected to detail texture extraction to obtain a texture feature map. Based on the texture feature map, the edge enhancement sub-image is subjected to texture fusion processing to obtain a fused sub-image, and the fused sub-image and the enhanced contrast map are then stitched together to obtain a sharpened target image.

6. The nighttime infrared surveillance imaging enhancement method according to claim 1, characterized in that, The step of performing feature annotation processing based on the clear target contour information to obtain a target feature annotation map includes: The contour feature points of the clear target contour information are extracted to obtain the coordinates of the feature points, and the geometric parameters of the feature point coordinates are calculated to obtain the target geometric parameter table; Based on the target geometric parameter table, feature regions in the clear target contour information are divided to obtain feature sub-region maps, and gray-scale feature statistics are performed on the feature sub-region maps to obtain sub-region gray-scale tables. Based on the grayscale table of the sub-region, key features in the feature sub-region map are identified, and feature attributes are labeled for the key features to obtain a feature attribute table. Based on the feature attribute table, key features are visualized and overlaid to obtain an overlaid feature map. The overlaid feature map is then integrated with annotation information to obtain a target feature annotation map.

7. The nighttime infrared surveillance imaging enhancement method according to claim 1, characterized in that, The process of performing background weakening processing on the target feature annotation map to obtain an enhanced imaging map includes: A mask is generated for the target region in the target feature annotation map to obtain a target mask map, and the target mask map and the target feature annotation map are superimposed pixel by pixel to obtain a target preserved map; The background area in the target retained image is grayscale suppressed to obtain a background suppressed image, and the background suppressed image is Gaussian blurred to obtain a blurred background image. Image fusion is performed based on the target-preserved image and the blurred background image to obtain an initial enhanced image. Then, edge transition optimization is performed on the initial enhanced image to obtain an enhanced imaging image.

8. A nighttime infrared surveillance imaging enhancement device, characterized in that, include: The acquisition module is used to acquire raw infrared images of nighttime scenes through infrared monitoring equipment, perform noise suppression processing on the raw infrared images to obtain a denoised infrared image, and perform contrast enhancement processing on the denoised infrared image to obtain an enhanced contrast image. The positioning module is used to locate the target based on the enhanced contrast map to obtain an initial positioning target box, and to perform regional grayscale analysis on the initial positioning target box to obtain an effective target box; The sharpening module is used to perform detail sharpening processing on the enhanced contrast image based on the effective target box to obtain a sharpened target image, and to perform contour extraction processing on the sharpened target image to obtain clear target contour information; The annotation module is used to perform feature annotation processing based on the clear target contour information to obtain a target feature annotation map, and to perform background weakening processing on the target feature annotation map to obtain an enhanced imaging map.

9. A computer device comprising a memory and a processor, wherein the memory stores 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 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images