结合天空区域分割的图像去雾算法  

Image dehazing algorithm based on sky region segmentation

作　　者：黄治勇 李震 李良荣 HUANG Zhiyong;LI Zhen;LI Liangrong(College of Big Data and Information Engineering,Guizhou University,Guiyang 550025,China)

机构地区：[1]贵州大学大数据与信息工程学院,贵阳550025

出　　处：《智能计算机与应用》2021年第7期13-19,共7页Intelligent Computer and Applications

基　　金：国家自然科学基金(61361012)。

摘　　要：针对现有去雾算法在天空区域易失真和复原图像整体偏暗的问题,提出一种结合天空区域分割的图像去雾算法。该算法首先利用图像局部香农熵将原图像分割为天空区域与非天空区域;然后对相应的透射率分别进行求取,并利用快速引导滤波对最终的透射率进行细化处理,再采用四叉树搜索算法求取大气光值,最后利用暗原色先验模型恢复出无雾图像;另外对于整体偏暗的复原图像,则利用改进的局部对比度保留的非线性增强方法进行亮度调整。实验结果表明,与其它去雾算法相比,所提算法能更有效地复原图像;复原后的图像无颜色失真和光晕效应,而且清晰度和亮度更佳,整体视觉效果更加符合人眼视觉特性。Aiming at the problem that the existing defogging algorithm is invalid in the sky area and the restored image is overall dark,this paper proposes an image defogging algorithm combined with sky area segmentation.The algorithm first uses the local Shannon entropy of the image to divide the original image into sky and non-sky regions;then calculates their transmittances separately,and uses fast guided filtering to refine the final transmittance,and then uses the quad-tree search algorithm to obtain the atmospheric light value,finally uses the dark channel prior model to restore the fog-free image;in addition,for the overall dark restored image,the improved local contrast preservation nonlinear enhancement method is used to adjust the brightness.The experimental results show that compared with other dehazing algorithms,the proposed algorithm can restore the image more effectively;the restored image has no color distortion and halo effect,has better clarity and brightness,and the overall visual effect is more in line with human vision characteristic.

关 键 词：图像去雾 局部香农熵 快速引导滤波 四叉树搜索 暗原色先验 亮度调整 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]