High performance pose of human face recognition for different color channels

This paper proposes a high performance pose face recognition system. A pre-processing phase is employed to isolate the faces from the background to show the robustness of the proposed system in changing background conditions. We compared between color histogram and Discrete Cosine Transform (D.C.T) method as two feature extraction methods. The color histogram performs face recognition by matching cross correlating of the histogram of a given face and that of the faces in the database. In DCT, we compared the feature vector of the selected image with the feature vectors of other faces, under the same nominal size, orientation, position, and illumination conditions using a Euclidean distance nearest-neighborhood method. The proposed system using color histogram achieves 99.8% recognition rate on face database of 150 subjects with varying poses. This rate is down to 93.3% for Discrete Cosine Transform (D.C.T).