Four Yuan Color Gabor Feature-based Face Recognition Technology Research

For decades, face recognition technology has been an active topic in image processing and pattern recognition. Its significance lies on both theoretical research and practical application. The Gabor wavelets, whose kernels are similar to the 2D receptive field profiles of the mammalian cortical simple cells, exhibit desirable characteristics of spatial locality and orientation selectivity. As a kind of transformation tool, Gabor transform is very potential in face recognition.This dissertation focuses on face recognition exploiting Gabor transforms. Because it has been mainly applicable to gray images, this dissertation presents a face recognition method based on color images, using quaternionic color Gabor features. Firstly, the color pixel is presented in the form of quaternion, and the complex Gabor transform is extended to quaternion space, so the convolution of this two is used for extracting quaternionic color Gabor features which can describe the color texture; secondly, in order to improve the algorithmic efficiency, the principal components analysis is used for the dimensions reduction of quaternion Gabor features; lastly, according to the specialty of the chosen sample sets, support vector machine which is suitable for the statistical learning and classification of small samples size but high feature dimensions is chosen to be the classifier.The face recognition method based on quaternionic color Gabor features presented by this dissertation combines color with texture - two basic image features. The experiments on Color FERET Database shows that these features including color information can not only reveal the differences among different face images but also have higher recognition rate. And the researches carried out on Lab color space prove that quaternionic color Gabor features is immune to illumination.