基于简化局部二元法的人脸特征提取  

Facial Feature Extraction Based on Simplified Local Binary Patterns

作　　者：袁健[1] 姚明海[1] 

机构地区：[1]浙江工业大学信息工程学院,浙江杭州310014

出　　处：《计算机技术与发展》2009年第6期84-86,90,共4页Computer Technology and Development

基　　金：浙江省自然科学基金(Y1080734)

摘　　要：提出了一种在复杂光照条件下,基于简化的LBP处理的人脸识别算法。局部二元模式(LBP)作为一种非线性高通滤波器,能够进行图像的边缘增强处理。由于噪声在任意方向上出现的随机性,LBP方法在边缘增强的同时,也对噪声做了增强,尤其是邻域内指数的权重系数对噪声的影响更加明显。基于此,对LBP算法进行了改进,提出基于相似度的人脸识别算法,实验证明算法有效地减少了因光照变化而产生的影响,同时缩短了运行时间,提高了图像识别率。A method based on the simple - LBP is presented for face recognition, which can work under the complex light conditions. Local binary pattern （LISP） filter can be used to the original image preprocessing as a nonlinear high- pass. As the noise appear randomly in any direction, LBP in the method of edge enhancement, noise also made to strengthen, in particular the exponential weights in the neighbors make the effect of the noises more prominent. For this reason,have improved LBP algorithm and presented the method based on the like- lihood- ratio classifier. Experimental method proved effective in reducing the impact as a result of the light changes, at the same time shorten the travel time,increased the rate of image recognition.

关 键 词：人脸识别 局部二元模式 光照归一化 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]