Color Face Recognition Based on Color Space Normalization and Quaternion Matrix Representation

The color face recognition method using quaternion matrix representation is based on RGB color space which is a fundamental and widely used space. However, the RGB color space is a relatively weak color space for face recognition due to the strong correlation between the three color channels. A technique named color space normalization (CSN) technique can reduce the correlation between the three channels of the RGB color space, and enhance the discriminating power of the normalized color space RGB-N. Therefore, in this paper, a method based on normalized color space RGB-N and quaternion matrix representation is firstly proposed. The proposed method not only has the advantage of quaternion matrix representation model, but also has the strong discriminating power due to CSN. The experiments in this paper are performed on the FERET color face database. Compared with the method based on RGB color space and quaternion matrix representation, the proposed approach in this paper can enhance the recognition rate.