Robust Face Recognition Method using AAM and Gabor Feature Vectors

摘要

The invention proposes a face recognition method using AAM (Active Appearance Model) and Gabor feature vectors. Gabor feature vector has the shape of the image, the scale, that the robust against small changes in rotation is well known for many uses as a feature vector of an object recognition. Away of a typical face recognition algorithm that uses the feature vectors EBGM (Elastic Bunch Graph Matching) requires the detection of facial features to extract the Gabor feature vector. However, the face feature point detection method is used in which EBGM is based on a Gabor jet similarity to, and sensitive to the initial point, detecting incorrect feature point affects the face recognition. On the other hand, AAM is known to be effective in the face feature point detection. The present invention proposes a method for detecting facial features and this approximate estimate based face recognition method, and face feature points by AAM estimated feature points to the initial point to refine the feature points detected by the feature point detection method based on Gabor jet similarity. ; that it is more robust than the detection method based solely on AAM and Gabor jet similarity-based facial feature detection method using Gabor jet similarity in facial recognition method using AAM and Gabor feature vectors provided by the present invention confirmed by experiments.