摘要
光照是人脸图像的成像条件之一,在人脸成像过程中具有很大的灵活性。由于拍摄照片过程中光照条件或观察角度的不同,导致人脸识别算法的性能明显下降。基于传统的区域校正思想提出了一种自动分块亮度校正算法,自动提取过亮块和过暗块并对其进行亮度校正。以彩色图像人脸数据库为例,快速、准确地得到亮度校正图像。仿真实验表明,将该算法校正恢复出的图像用于人脸识别,可以提高人脸识别率,具有一定的应用价值。
llumination, as one of the conditions of image formation, plays a flexible role in the process of human face imagery. Due to the different conditions and angles in the photograph shooting, the performance of skin-color detection and face recognition will be remarkably decreased. An algorithm of automatic partitioning blocks illumination correction is proposed based on the traditional region correction. The whole image is divided to various needs. Also, the illumination correction can be rectified if necessary. The illumination correction image is gained accurately using the color facial image database in the paper. The simulation experiment shows that it can improve the percents of face recognition using the corrected image. The algorithm has further application value for face recognition.
出处
《计算机仿真》
CSCD
2008年第5期185-189,共5页
Computer Simulation
基金
山西省自然科学基金项目资助(2006011028)
国家自然基金项目资助(70471003)
关键词
亮度直方图
亮度校正
回归分析
Illumination histograms
Illumination correction
Regression analysis
