Faces Feature Extraction And Recognition

Face recognition in recent years is one of research focus, which is used in pattern recognition, image processing, machine vision, computer graphics technology, science and the field of biometrics. Face recognition's purpose is to identify the input image, which compare the input image and test image.Face recognition technology is developed from the front of gray face image in a simple background to multi-profile (front, side, etc.) face in complex background. Now face recognition is developing to 3D. In the course of its evolution, face recognition technology became more complex. Recognition results is continue to be improved. But now face recognition is not effective in complex background. Therefore, this article is basis on previous research, searching and analysis of a large number of domestic and foreign academic works on face recognition. Fouce on the core technology and algorithms of facial feature extraction and recognition. The purpose is to seek for face recognition algorithms which is relatively simple, feature extraction and recognition quick, high recognition rate.First, the article describes the research background and significance of the face recognition technology and development process in the domestic and international. Form face recognition technology perspective, introduce face detection, feature extraction and classification techniques in categories.Second, use wavelet transform and BPCA (Block Principal Component Analysis, PCA) to extract facial feature and recognition. The method has higer normalization for the input face image, and is vulnerable to the impact of changes in illumination and pose. Therefore, this paper uses histogram equalization technique to preprocess the face images to eliminate the impact of differences in light intensity; Secondly, use wavelet transform to extract face images of relatively stable low-frequency sub-band, blurring the facial expression and posture influence, and also reach to reduce vector dimension. Use BPCA to feature extraction, principal component analysis algorithm itself is based on the overall gray face image to extract feature vector correlation. On this basis, use BPCA to extract the feature, according to sub-block contribution to the recognition results generated weighting factors.Get integrated projection coefficient, and finally use the nearest neighbor classifier to classify.BPCA algorithm is better than the traditional method. It greatly reduces the computational complexity, the extracted features further reflects the difference between the face and improve the recognition rate. Experimental results show that the algorithm used in this paper is accurate and effective.