基于图匹配的人脸识别算法

为了识别有较大表情变化的人脸,利用sift算法找到人脸图像中的DoG关键点,用这些极值点所包含的灰度信息和位置信息建立最小生成树,再使用最小生成树估计联合Rényi熵,结合图论中的图匹配知识,进行人脸识别.实验结果表明,该算法在人脸出现较大表情变化的情形下仍能够得到较为准确的识别结果.

The face recognition algorithm based on graph matching

To recognize a human face when the greater changes in the expression of the human face arise,a new algorithm is given.Using sift algorithm to find the DoG points in a face image that is invariant geometric features.Minimum spanning tree is established with grayscale infor-mation and location information corresponding to these extreme points,then using the mini-mum spanning tree to estimate joint Rényi entropy,finally combined with knowledge of graph matching in graph theory,recognition of human face is performed.Experimental results show that the algorithm can deal with the situation that a larger change in the facial expression better and get more accurate results.