Face Recognition Algorithm Study Based On Edge Information

Face recognition is one challenging subject in biologic recognition field, which involves many domains such as feature abstract, computer vision, image processing, pattern recognition and identity authentication et al. There are many algorithms about face recognition now, but face recognition employing the spatial information of edge map associated with optimum information remains an unexplored area. According to synthetically analyzing present condition, the basic theory and method of edge information used in face recognition is studied in this paper.Edge information is useful object feature insensitive to illumination changes to certain extent. In this paper, a new modified Hausdorff distance measure using local distance mapping and a recognition strategy using "deputy face" to validate discriminant power are proposed, which uses local distance mapping result as feature replacing edge map, the most eigenface combining edge information, respectively.The per-processing and several illumination compensation methods are introduced firstly. It is pointed that surface fitting to edges can not only effectively reduce illumination influence but also hold dominant parts' effective information, and offer reliance for face recognition. At last, a new modified Hausdorff distance measure strategy is proposed using local distance mapping as feature. And the face recognition system is realized by Visual C++ programming under Windows operating system.Experimental results show that the modified algorithm can achieve recognition rate of 80.4% based on 156 faces in ORL database. It is obvious that this algorithm can not only achieve better recognition rate but also emphasize the importance of the crucial features. The proposed idea provides a new way for face recognition.