结合天空分割修正的快速去雾方法  

Fast Dehaze Algorithm Based on Sky Region Segmentation and Modification

作　　者：刘子兵 戴声奎[1] 

机构地区：[1]华侨大学信息科学与工程学院,福建厦门361021

出　　处：《华侨大学学报（自然科学版）》2018年第1期133-138,共6页Journal of Huaqiao University(Natural Science)

基　　金：中央高校基本科研专项基金资助项目(JB-ZR1145);福建省科技计划重点资助项目(2013H0030)

摘　　要：针对目前去雾算法对高亮天空区域处理不理想,以及去雾后的图像整体视觉效果较差的问题,提出一种结合天空区域分割修正的快速雾天图像复原方法.首先,对输入图像进行白平衡处理;其次,根据大气散射物理特性和光学成像特性对大气耗散函数做初始估计,判断是否存在天空区域,若存在天空区域,结合对比度增强调整,Otsu算法(大津法)分割出天空区域,修正天空区域的大气耗散函数;最后,由大气散射模型得到复原图像,并对复原图像做亮度调整.实验结果表明:该算法具有较强的场景适应能力,能很好地处理天空区域;复原图像具有较好视觉效果,而且执行速度更快.In order to overcome the existing dehaze algorithms defects,such as bad effects in bright sky area and poor visual effects to dehaze images,a fast dehaze method is proposed based on sky region segmentation and modification.Firstly,the white balance processing is used for input images.Secondly,based on the physical property of the atmospheric scattering model and optical imaging,atmospheric veil is roughly estimated,and then determines whether or not there exists sky region.If sky region exists,sky region is segmented by contrast adjustment procedure and the Otsu algorithm,and then atmospheric veil of the sky region is modified properly.Finally,haze-free images are restored by atmospheric scattering model,and a post-processing procedure is adopted to adjust the brightness of the restored images.Experimental results yield that the proposed method not only has robust scene adaptability and achieves good restore performance in sky area,but also maintains good visual effects and ensures the faster performance.

关 键 词：图像复原 去雾 图像分割 大气散射模型 大气耗散函数 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]