基于鲁棒估计的Retinex图像增强方法  

Retinex image enhancement method based on robust estimation

作　　者：王晓莉[1] 文华[2] 罗予频[1] 

机构地区：[1]清华大学自动化系,北京100084 [2]长沙理工大学设计艺术学院,湖南长沙510015

出　　处：《计算机应用》2006年第7期1611-1613,共3页journal of Computer Applications

摘　　要：针对原变分Retinex方法的人工伪影问题,指出理想重建光照应该具有分区域恒常的特性,提出一种以阴影边界作为出格点、基于鲁棒估计的变分Retinex图像增强方法。用合成图像与实际图像进行实验表明,本改进方法不仅可以有效压缩图像动态范围,显示淹没在阴影区域的细节,而且解决了原Retinex方法的伪影和扭曲等问题。Because of failing to reconstruct piecewise constant illumination, variation Retinex method suffers from artificial halos and color distortion. A Retinex method based on robust estimation was proposed, which took shadow edges as outliers and smoothed range by range. Application of method to synthetic image and natural image proves that the improved method preserves high dynamic range compression and avoids halo effect and color distortion.

关 键 词：Retinex方法 鲁棒估计 变分方法 动态范围压缩 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]