Color Face Detection And Recognition Research

It is well known that face detection and recognition has been active in the forefront of the research area of pattern recognition. Looking for a more effective method for face detection and recognition is still a challenging problem.The real world is colorful. Recent researches efforts show that color play a very important role in face detection and recognition. The face detection and recognition based on color images achieves better performance than face detection and recognition based gray or illuminant images. Thus, in this thesis, we did researches on face detection and recognition based on color images. The main work of this paper is summarized as follows:First, we analyzed and summarized the development and the basic methods of face detection and recognition.Second, we analyzed the color spaces in which the skin model build and the color face detection based on skin model. We applied the Fisher discriminant analysis to skin color and non skin color segmentation. We proposed an appropriate color space for face detection based skin color, named as "Skin Disciminant Color Space (SDCS)". Experiments show that in SDCS color space, skin color is more separable from non skin color.Third, we analyzed the theory of Gabor wavelet transformation and its parameter meaning. Then, we introduction the face representation method based on Gabor features. As Gabor filter with different parameter has different ability in face representation, we proposed a novel face representation method named "Weighting Gabor wavelet". We used experiments by ORL, FERET and AR face database demonstrate that our method is better than direct Gabor wavelet face representation.Forth, we analyzed the color spaces used in face recognition. Then, we proposed a novel face recognition method based multiple features in the color space. Finally, we demonstrate the effectiveness of our method by experimenting on AR and NUST_RWFED color face databases.