未知光源参数的人脸光照恢复方法  

Restoration of Human Face Illumination with Unknown Light Source

作　　者：张智斌[1] 赖剑煌[1] 邓东皋[1] 

机构地区：[1]中山大学计算机视觉研究中心

出　　处：《计算机研究与发展》2005年第4期629-634,共6页Journal of Computer Research and Development

基　　金：国家自然科学基金项目(60144001);教育部留学回国人员科研启动基金项目;广东省自然科学基金项目(021766)

摘　　要：数字人像技术在实际应用中有广泛的应用前景.在未知光源位置的非均匀光照射下的人脸光照归一化问题是人脸识别技术的一大难题.对在未知光源位置的非均匀光照射下的正面人脸图像提出了一种基于参考人脸模型的方法.它使用参考人脸模型估计人脸的光照阴影区域,把人脸的纹理表示为训练图像模型的线性组合,根据人脸光照明亮区域的纹理信息,使用最小二乘法估计出最优的线性组合系数,从而重建原图像人脸纹理实现人脸光照恢复.实验表明这种方法对于阴影处的人脸信息恢复不仅是有效的而且是快速的.Human face recognition is an important technique that can be widely employed in many practical applications, and, specifically, recognizing human face under non-uniform illumination condition with unknown-position light source is a particularly difficult problem. In this paper, a human-face-model based approach to restore the illumination of human face under the above condition is proposed, which work by recovering the intensity of the shaded areas in human face. Human face texture is represented as a linear combination of several training images, and, based on the face texture in the bright area, face illumination is restored by recovering face texture through a least-square estimation. Experiments show that the method is not only efficient but also fast in restoring the shaded area on human face.

关 键 词：参考人脸模型 光照归一 PCA方法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]