Multi-view Face Recognition with Unified Subspace Analysis

• Main Authors: Ching-chih Cheng, 程敬智
• Other Authors: Kuo-Chin Fan
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/31559620622515797240

碩士 === 國立中央大學 === 資訊工程研究所 === 98 === Face recognition technique has been developed for several years. The research results can be found in several applications, such as airport security system, access control system, ATM verification system, and surveillance system. However, the performance of identification result will be heavily affected by the factors of expression, pose, illumination and resolution of image. Hence, many relating algorithms were developed focusing on resolving these problems. Until now, most of the existing algorithms can still not fully resolve the pose problem in face recognition. In response to this need, we present a two stage identification method to improve the performance of face recognition system handling pose problem. In our proposed method, we first predict the pose variations for all input images. After that, we further classify the image class in the corresponding pose label set. The advantage of our proposed method can not only eliminate the difference in different pose sets but also reduce image matching time for recognition. Experimental results reveal that the proposed face recognition method with pose classifier can achieve better classification result.