Face Recognition by Geometrical Feature-Point Bilateral Matching
Yea Shuan Huang, Guo Da Peng · Applied Mechanics and Materials · 2013
This paper presents a novel feature-point bilateral recognition method for recognizing human faces. At first, from either an input face image or a reference face image, a set of distinct feature points is extracted by using a general salient point detection algorithm. Then, based on the detected feature points, a bilateral recognition is performed. Bilateral recognition means there are two ways of recognition, forward recognition and backward recognition. Finally, the forward score and the backward score are summed up into a bilateral score which is used to obtain recognition result. In order to perform recognition in real-time, we also use a GDA algorithm to select the possible candidates, and then use the proposed bilateral recognition operation to make the final recognition decision. Experiments on two famous face databases show that the proposed algorithm get excellence recognition result and is complementary to traditional global-feature-based face recognition methods.