Bayesian-based Face Recognition

Face recognition more and more becomes a research hot issue because of its extensive society application prospect. It is a multi-subject cross issue concerning on image process and pattern recognition and physiology and psychology etc. Human recognition itself is a very easy thing even people' mentality can be estimated according to facial expression.But it is difficult for machine to recognize face automatically which needs studying from all aspects. The main procedure of face recognition automatically is the pre-process of image and the feature abstraction and the classification, and among them the image pre-process is the precondition of the feature abstraction and the classification as well as the key problem which affect the rate of right recognition fate. The purpose of image pre-process is to make the face image with the consolidated dimension and to eliminate the impact of lighting and gesture for the recognition rate . The precondition of image pre-process is the location of the feature spot especially the eyes location automatically.The article proposes a methodology of eye location automatically based on the step direction and template matching and carry out the preprocess to image by means of the result of this method. The image through preprocess are used to recognize with the principal component analysis and Bayesian method. Eventually a picture comparison face recognition system is designed according to the above theory.The eye location method automatically of the article is based on the digit image process. Twenty percent of face image of eye status is close or open little in accordance with experience knowledge and statistics result. So the article adopt different method to locate the eye on the grounds of the different feature with eye status. The procedure of eye location of the article is that first the eye area can be conformed take advantage of integral project curve and covariance project curve and then the binary edge image of the area can be obtained with the edge operator. If the half-circle outline can be detected and then according the feature of the step direction emitting forward we can verify whether the half-circle is eyeball. A step eye template is used to matching the eye area for that image with which the eyeball can not be located, thereby the eye position can be found. The images can be normalized according to the eye location. The...