基于视觉区域划分的雾天图像清晰化方法  

Foggy image enhancement based on regions of human visual

作　　者：张新龙[1] 汪荣贵[1] 张璇[1] 朱静[1] 

机构地区：[1]合肥工业大学计算机与信息学院,合肥230009

出　　处：《电子测量与仪器学报》2010年第8期754-762,共9页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金(编号:60705015)资助项目;国家自然科学基金(编号:60575023)资助项目;安徽省自然科学基金(编号:070412054)资助项目

摘　　要：针对雾天图像退化程度与景深呈非线性关系的特点,以及MSR算法的参数选择没有考虑图像景深信息的局限性,提出一种新的基于局部信息特征的自适应Retinex雾天图像增强算法。算法将雾天图像划分为不同的子块,根据子块像素所处的人类视觉区域反映出的雾的厚薄程度自动调整Retinex增强算法的参数取值,使用面积比例因子策略融合处理后子块以避免分块带来的块状效应,最终实现对雾天图像的清晰化。实验表明:本文方法比MSR算法在多景深雾天图像局部细节增强方面具有更好的效果。The degradation of foggy images is non-linear with the depth of field,and MSR algorithm does not consider the depth information of field in images.A new adaptive Retinex algorithm in foggy image enhancement based on local area information is proposed.The algorithm divides foggy images into different blocks,and then adjusts the parameters of Retinex automatically according to the strong degree of fog reflected by the region of human visual response which the block belongs to.The block artifacts will be avoided from block dividing by an area rate factor strategy combined with the processed blocks.The subjective observation and objective evaluation show that,the proposed method is more effective than MSR algorithm in local details enhancement of foggy images with multi-depths of field.

关 键 词：计算机应用 雾天图像增强 人类视觉模型 RETINEX 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]