块匹配算法在基于MPEG-4的人脸特征提取中的改良及应用  

Improvement and Application of BMA in Facial Feature Extraction Based on MPEG-4

作　　者：王雁来[1] 尹宝才[1] 孔德慧[1] 谷春亮[1] 

机构地区：[1]北京工业大学计算机学院多媒体与智能软件技术北京市重点实验室,北京100022

出　　处：《北京工业大学学报》2003年第1期98-102,共5页Journal of Beijing University of Technology

基　　金：国家"973"基金资助项目(2001CCA03300);国家"863"基金资助项目(2001AA114160);北京市教委基金资助项目(P070702-01);北京工业大学研究生科技基金资助项目(YKJ2002-14)

摘　　要：首先对视频内容进行了分析,针对不同人脸的光照效果和色度变换效果做了平滑处理,以减少灰度突变区域。然后运用匹配算法进行像素匹配。为减小计算量,添加了特征点位置约束关系和梯度约束关系的限制,同时改进了相关系数的计算方法,这样不仅有效提高了匹配准确度,同时还减少了计算复杂度。试验证明匹配效果显著,而且计算量没有增加。The block match algorithm (BMA) can efficiently extract facial definition parameter (FDP) and facial animation parameter (FAP) in human face video which is based on MPEG-4. Firstly, video frequency is analyzed and video illumination is smoothed so as to reduce luminance mutation. Then the block match algorithm is used to do the match of picture element. For the purpose of decreasing computation, the limitation for the position and gradient relationship of those feature points is appended. Furthermore, in order to get an exact match result, the correlation coefficient is improved. The effect of improved algorithm for experimental results is prominent

关 键 词：块匹配算法 MPEG-4 人脸图像 特征提取 图像处理 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] Q983.1[自动化与计算机技术—计算机科学与技术]