Face Recognition Using Independent Component Analysis of Gabor Responses

This paper addresses new face recognition method based on Independent Component Analysis(1CA) and Gabor filter. Our method consists of three parts. The first part is Gabor filtering on predefined fiducial points that could represent robust facial features from original face image. The second part is transforming the facial features into the basis space of ICA, which is able to represent individual facial features optimally. Thus, trained face images are represented as points in the space. In order to identify, test images are also projected into the basis space of ICA from image space and compared to the training images using fisher linear discriminant(FLD) in the space. The basic idea of combining ICA and Gabor filter is to overcome the shortcomings of ICA. When raw images were used as input of ICA, the basis space of ICA cannot reflect the correlation of facial feature well, because original face images have deformation due to in-plane, in-depth rotation and brightness and contrast variation. So, we have overcome these problems using Gabor filter responses as input data. Gabor filter has the robust characteristics in illumination and rotation. Four face recognition method PCA, Gabor filter response, PCA of Gabor filter response and ICA are used in the recognition experiments. We confirmed the improvement of discrimination ability when the Gabor responses had transferred to the space constructed by the independent components. And, our method has excessive advantage in gallery DB size than recognition method only using Gabor filter responses.