Human Face Recognition Research And Application Based On Lightness Normalization

Human face recognition is a relatively difficulty problem up till now. Spe-cialy for problems about uneven illumination and instability in most human face recognition system. But for a stability system of human face recogniation these problems must be solved, and these problems are key elements in recognition sytem. This paper proposes take advantage transfer between color space related equipment RGB and color space unrelated equipment L*a*b*. After that, mean-while make use of normalization of light from non-standard to standard in nature. According to this method, images have even illumination among every parts of image. So ratio of recognition has been improved by the way. Because reserrch of human face recognition is an important topic in the area of pattern recognition and artificial intelligent, and it has so many and broad application.This paper pro-posed the method is that based on wavelet decomposition analysis, and this one is mainly used into extract characteristic of human face. Because this one dramat-ically reduced space of storage of images and reduced complexity of computing. The experiment of this paper is divied into two stage, one is preprocedure which include color space transfer and even illumination, last is coner point checking based on Harris algorithm. The arm of using Harris algorithm is for searching eye position in human face image and then realize geometry normalization. Then in first stage, key procedure is taking advantage of Sym8 serial wavelet in order to compress image. The second stage is using BP neural network to realize the training of human face images and simulation of human face images. Last, this paper will conclude that ratio of using illumination will be improved to 87.6% compared to the ratio of non-using illumination is only 74.2%.