作者: Yongyue Zhang , Zhenyun Peng , Suya You , Guangyou Xu
DOI: 10.1007/BF02943172
关键词:
摘要: In many automatic face recognition systems, posture constraining is a key factor preventing them from application. this paper, series of strategies-will be described to achieve system which enables under varying pose. These approaches include the multi-view modeling, threshold image based feature detection, affine transformation normalization and template matching identification. Combining all these strategies, with pose invariance designed successfully. Using 75MHZ Pentium PC database 75 individuals, 15 images for each person, 225 test various postures, very good rate 96.89% obtained.