摘要
针对五官的局部特征,提出一种具体的识别方案。与传统弹性图匹配的方法相比,该方案提取的人脸局部的特征,大大降低了运算量,理论和实验都表明了该方案的可行性。在对人脸图像预处理之后,利用积分投影的方法,获取五官的位置,而后采用奇异值分解(SVD)去提取五官附近的局部特征,作为人脸的主要特征。比较了SVD和离散傅里叶变换(DFT)两种特征提取的方法,结果表明,通过SVD提取人脸的局部特征在减少计算量的基础上,能够很好保留人脸的特征,达到了较好的识别效果。
A face recognition method based on the local facial features is presented in this paper. Compared with the traditional elastic matching, this method can effectively reduce time complexity. Firstly, the edge information of face is acquired, and then according to that, the positions of facial features are detected by integral projection; Secondly, according to those posifiom, the local area of facial feature can be obtained;Thirdly,the featurres of local images will be extracted by SVD as the main features of a human face. Experimental results show that the features picked up by SVD can well represent the human face and the recognition results are satisfactory.
出处
《济南大学学报（自然科学版）》
CAS
2007年第1期56-59,共4页
Journal of University of Jinan(Science and Technology)
基金
山东省高等学校中青年学术骨干计划(Y0204)
济南大学科技计划(Y0521)
关键词
人脸识别
积分投影
局部特征
奇异值分解
face recognition
integral projection
local character
singnlar value decomposition
