基于光学深度估计的雾天降质图像清晰化复原  

Fog-degraded Image Clearness Restoration Based on Optics Depth Estimation

作　　者：翟艺书[1] 王宏[2] 

机构地区：[1]天津工程师范学院理学院,天津300222 [2]河北理工大学理学院,河北唐山063009

出　　处：《计算机仿真》2010年第3期250-253,共4页Computer Simulation

基　　金：天津市自然科学基金项目(08JCYBJC12100);天津工程师范学院自然科学基金项目(KYQD08011)

摘　　要：研究雾天图像质量优化问题,雾天拍摄的户外场景图像对比度和色彩都发生不同程度的降质,严重影响了视觉监控系统的可靠性和鲁棒性。为提高雾天的图像清晰度,根据大气散射模型和雾天图像的对比度衰减同场景点深度成指数规律的先验知识,提出了一种场景光学深度的估计方法,应用于雾天降质图像的清晰化复原。采用建立了两组联系大气散射系数与场景点深度的关系式,并计算出场景各点光学深度的估计值。同时结合了两幅差异雾况天气下同场景图像的灰度信息以及局部对比度信息,可以更好地提高对场景深度信息估计的可靠性。结果表明,方法获得的场景可以实现雾天降质图像对比度的有效改善,获得满意的视觉效果。Images captured in fog weather suffer from poor contrast and color, which seriously degrade the reliability and stability of outdoor surveillance system. According to the apriori knowledge of atmospherical seattering model and the contrast exponential degrade law, this paper proposed a new optics depth estimation method, which was further applied to restoring fog - degraded images. Firstly, this method constructed two equations about atmospherical scattering coefficient and depth of each pixel in scene image, and then, the optics depth estimation value for each pix- el was computed out. Since the method reasonably combines the gray information and local contrast information from two degraded images in different weather, it can improve the reliability for depth information estimation. Experimental results demonstrate its satisfying scene restoration performance.

关 键 词：大气散射模型 雾天降质图像 光学深度估计 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]