Abstract
This paper presents an improvement of our previous research that applies color information in order to improve the recognition rate and equal error rate (EER). Our previous research just considered the luminance (Y) element in facial features extraction. In this research, we also consider the chrominance elements (Cb and Cr) to deal with the face skin information. The facial features vector is extracted using non-block DCT with selecting just small part of dominant frequency elements. Experimental results using data from four face databases containing 2268 images with 196 classes show the YCbCr color configuration improve the rank one of our previous result by about 5.91%. However, it requires three times more of the previous method time consumptions.
