Research On Face Recognition And Gender Recognition Algorithm Based On Multi-scale Imageanalysis

Through the facial recognition technology for identity authentication is a research and application hotspot. Using the computer we extract and analysis face feature to realize the recognition of people as well as relevant classification of facial information such as gender identification.In this paper, while we extract the features of faces, at first we owe the sub-graphs of face image by wavelet decomposition. We make full use of texture characteristics of high frequency sub-graph which affected by the light small. Using LBP extended operator we account statistical histogram of each sub-graphs multiplied by the corresponding weights according to the contribution of each sub-graphs. The characteristic vectors of face image that we cascade the statistical histograms are used to overcome the interference of illumination.When we extract the features of face for gender recognition, we use the first and second order low-frequency sub-graph of face after the face pictures are decomposed with wavelet. The statistical histograms of the sub-graphs are obtained by using LBP operator arithmetic. In the same way, we owe the LBP histogram statistics of the primary face picture. Then we cascade all the statistical histograms as the classification feature vector of face gender recognition.This method combined with the feature of the integral structures and local details better. It improves the stability of classification and solves face gender recognition based on small sample training under the library. And this method reduces computing and storage overhead.It can satisfy the requirement for storage space and computational complexity on the mobile and embedded platform.