提出一种非约束条件下的人脸识别方法。利用Prewitte算子将人脸图像转换成特征图像以保护更多的局部信息,并将特征图像划分为多个区域特征向量表示人脸,通过调整二次直方图距离中的权系数矩阵以降低干扰因素的影响,利用改进的距离测度计算特征图像间的相似度。实验结果表明该算法与传统的识别方法相比,有一定的人脸特征描述和识别性能。%A face recognition method under no constraint conditions is proposed. Facial images are transferred to a set of block feature vectors using Prewitte operator, in which more local information is considered. A modified Quadratic Histo-gram Distance(QHD)is applied to measure the similarity of these feature vectors, the weight matrix of QHD is adjusted to improve recognition accuracy. Face recognition is implemented based on combination of the modified measurement and feature descriptor. Experimental results show the algorithm has better performance than original methods to some extent.
展开▼