基于自适应天空区域分割的图像去雾算法  

Image dehazing algorithm based on adaptive sky region segmentation

作　　者：邓翔宇 李俊腾 DENG Xiang-yu;LI Jun-teng(School of Physics and Electronic Engineering,Northwest Normal University,Lanzhou 730070,China)

机构地区：[1]西北师范大学物理与电子工程学院,甘肃兰州730070

出　　处：《计算机工程与设计》2024年第9期2742-2748,共7页Computer Engineering and Design

基　　金：国家自然科学基金项目(61961037);甘肃省高等学校产业支撑计划基金项目(2021CYZC-30)。

摘　　要：针对暗通道算法在处理图像中天空区域时会出现色彩失真和光晕现象以及大气光值错误估计的问题,提出一种改进的暗通道先验去雾算法。利用雾图中天空区域的物理特征粗略分割图像天空区域,从天空区域中对大气光值进行估计;设计一种自适应双阈值分割方法用于天空区域的精细分割,提出一种基于亮通道先验的透射率估计方法,将亮通道先验模型估计的天空区域透射率值和暗通道先验模型估计的前景区域透射率值进行融合;引用一种快速引导滤波器对融合后的透射率进行边缘细化,基于大气散射模型完成图像的复原。实验结果表明,该算法很好地克服了暗通道算法的缺陷,保留了复原图像的视觉真实性。In view of the problems of color distortion,halo phenomenon and wrong estimation of atmospheric light value when dark channel algorithm is used to process the sky region in the image,an improved dark channel prior dehazing algorithm was proposed.The sky region of an image was roughly segmented using the physical characteristics of the sky region in the haze image,and the atmospheric light value was estimated in the sky region.An adaptive dual-threshold segmentation algorithm was designed for fine segmentation of the sky region,and a transmission estimation method based on bright channel prior was proposed.The sky region transmission value estimated using the bright channel prior model and the foreground region transmission value estimated using the dark channel prior model were fused.A fast guided filter was cited to refine the fused transmission,and the image restoration was completed based on the atmospheric scattering model.Experimental results show that using this algorithm effectively overcomes the defect of the dark channel algorithm,and well preserves the visual authenticity of restored image.

关 键 词：图像去雾 天空区域分割 自适应双阈值分割 暗亮通道先验 大气光估计 透射率 大气散射模型 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]