基于小波商图像的人脸光照补偿  

Facial illumination compensation based on wavelet quotient image

作　　者：刘丽华[1] 王映辉[2] 邓方安[1] 

机构地区：[1]陕西理工学院数学系,陕西汉中723001 [2]西安理工大学计算机科学与工程学院,陕西西安710048

出　　处：《计算机工程与设计》2009年第14期3402-3405,共4页Computer Engineering and Design

基　　金：国家自然科学基金项目(70472072);陕西理工学院博士基金启动项目(SLGQD3030)

摘　　要：复杂光照条件下的人脸识别是一个困难但需迫切解决的问题,为此提出了一种有效的光照补偿算法。该方法根据人脸光照线性变换子空间理论——商图像理论,构造了小波低维训练集,实现了对待识别图像的光照条件估计,并且通过加光和去光策略增强了光照补偿效果。与传统商图像理论相比,该方法利用小波分解,提高的算法执行效率,实验结果表明,该算法以较小的代价取得了较高的识别性能。Face recognition under complex illumination conditions is still an difficult but must deal with problem an effective illumination compensation method is proposed. Based on quotient image theory, the illumination condition is estimated on facial wavelet de-dimension illumination training set. Moreover, the aim of facial illumination compensation is implemented by two basic strategies of adding light and reducing light. Compared with traditional quotient image theory, this method enhances arithmetic efficiency by wavelet transform. The experimental results show that the improved methods get a very competitive recognition rate with low computational cost.

关 键 词：人脸识别 商图像 光照补偿 小波变换 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]