融合遗传算法的多域值分块低照度图像增强算法

Multi-domain Block Low-illuminance Image Enhancement Algorithm Combined with Genetic Algorithm

针对低照度图像对比度低、细节模糊等问题,提出了融合遗传算法的多域值分块低照度图像增强算法。通过遗传算法寻找输入图像亮度通道的最优分割阈值,根据得到的阈值,将亮度通道划分为多个不同曝光级的子图。通过多阈值分块增强算法对所有子图进行评估,根据评估结果调整每个子图的亮度。最后使用多尺度融合方法将输入图像细节信息融合到亮度增强图像中与现有的增强算法进行对比实验。结果表明,用所提算法增强后图像的各指标增加幅度均大于其它对比算法,且在提升图像亮度的同时解决了增强图像颜色失真和亮度分块的问题,有效地还原了图像的纹理信息,增强后图像的亮度分布很好地还原了真实拍摄环境的亮度分布,验证了该算法具有更好的性能。

Image process is widely used in route planning,industrial damage detection,face recognition,medical aided diagnosis and other fields,and the vigorous development of this technology also has higher requirements for the performance of image acquisition equipment. The insufficient exposure and inconspicuous texture of the detected object in the image will be affected by the general deviation of the image quality collected by the low-end equipment,while the high-end image acquisition equipment is generally expensive and has a precise structure, which is not suitable for use in harsh acquisition environments. At the same time,the distance between the position where the detected object appears and the image sensor is also affected by uncontrollable factors such as randomness. These bad factors all increase the difficulty of the later target recognition task,and also bring a bad visual experience to the user.Therefore,it is of great theoretical significance and application value to design an enhancement algorithm to improve the quality of low-light images.To solve the problems of low contrast and blurred details in lowillumination images,a multi-domain block low-illumination image enhancement algorithm fused with genetic algorithm is proposed. The algorithm can be divided into four stages:color space conversion,brightness enhancement,detail enhancement and multi-scale fusion. First of all,to prevent the original color characteristics of the image from changing when the image brightness is enhanced,the input image is converted from the RGB color space to the HSV color space,so as to better separate the color information and brightness information of the input image,so that the color of the image can be improved. Information is not altered when augmented. When the algorithm enhances the image brightness,in order to prevent the over-exposure or under-exposure of some areas of the processed image,it is necessary to reduce the impact of the complex exposure of the acquisition scene on the enhanced image brightness. Therefore,the multi-threshold block enhancement is more in line with the actual scene. The method is used to enhance the brightness of the image. In order to improve the processing speed of the algorithm,the genetic algorithm is used to search for the optimal segmentation threshold of the brightness component of the input image.Then,the luminance channel of the input image is divided into a plurality of different exposure level subimages according to the obtained multiple thresholds and the image luminance gradient law. The detailed information contained in each sub-image is the evaluation criterion,and the complexity of all sub-images is evaluated through a multi-threshold block enhancement algorithm. The brightness of each sub-image is adjusted according to the evaluation results,and the arrangement order of the brightness of each sub-image after enhancement same as that before enhancement. To enhance the detail information of the image,the guided filtering algorithm is introduced,and the original image is subjected to two guided filtering processes. The first process filters out image noise,and the second filter enhances the contour information of the image. After two filtering processes,the enhancement results with rich contour information and less noise are obtained. The unsharp mask algorithm is introduced,which uses low-pass filtering to obtain the low-frequency information of the original image,subtracts the original image and the low-frequency information to obtain the high-frequency information of the image,and superimposes the enhanced highfrequency information with the original image to obtain the high-frequency information of the image.Finally,a multi-scale fusion algorithm is introduced to decompose the enhanced input image contour information,enhanced input image texture information details and brightness enhanced input image into Laplacian pyramid and Gaussian pyramid. The resulting Laplacian input and the corresponding Gaussian weight map at each level obtain the final enhancement result from texture information,contour information and brightness information. The algorithm is also compared with the existing enhanced algorithms. The results show that the increase of each index of the image enhanced by the proposed algorithm is greater than that of other comparison algorithms,and the proposed algorithm effectively solves the problem of color distortion and brightness blocking while enhancing image brightness,effectively restoring the texture information of the image. At the same time,the brightness distribution of the enhanced image restores the brightness of the real shooting environment. It proves that the algorithm has better performance.