The Research Of Face Recognitionbasedon Wavelet And Improved Epidemic Algorithms

As in recent years, the computer technology has a rapid development, the facerecognitiontechnology as an application of computer applicationis also in its research upsurge. The technology of facerecognition is a technology that combined new technology and people’s biometric characteristics, which is a hot research topic in image processing and artificial intelligence. Face recognition is more and more used in public security, identity authentication, human-computer interaction and so on.Generally speaking, face recognition is divided into five steps: face images acquisition and detection, face images pre-processing, face images feature extraction, face images classification. Local tangent space alignment algorithm is one of the most successful and advanced manifold algorithm at present domestic scholars.The wavelet transform is a tool that is often used in the image extraction, using wavelet transform to extract the feature information of the face image has the character of multi-scale, multi-direction, and a good spatial characters.In this paper, the author mainlystudies thesteps of face image extraction,proposing a face recognition model mixed the wavelet transform and improved local tangent space alignment algorithm.The algorithm is used to extract the main features of image information, and the identification of human face recognition results has improved significantly.The main work and innovations are as follows:(1) By the inspiration of Gabor wavelet transform and the double dual tree complex wavelet transform, this paper presents a wavelet transform based on maximum entropy to extract the feature information of face image.(2) The X-means algorithm adopted to divide the sample into some overlapping blocks and the number of the blocks is much smaller than the size of samples. And after this, we can obtain the sample blocks’ projection information in its local tangent space, and then integrated the local arrangement information into the global coordinate.(3) After the test samples’ wavelet transform we can obtain the feature information.We can reduce its dimension by local tangent space algorithm. And then getting the face recognitionclassification resultthrough the 3-neighborhoodalgorithm.The proposed face recognition model is tested on the ORL database contains 400 face images and the Yale database contains 165 face images, and compared with the three method of popular facerecognition models.Experimental results show thatthe face recognition algorithm of fused two kinds of wavelet and manifold algorithm not only can effectively extract the face feature information, but also can improve the accuracy.