A Novel Color Face Recognition with Semi-orthogonal MPCA Method

In this paper, the semi-orthogonal multi-linear principal component analysis (MPCA) method has been proposed for color face recognition. Recently, MPCA seems to be an appropriate scheme for dimensionality reduction and feature extraction from color images, handling the color channels in a natural, “holistic" manner. However, it is difficult to develop an effective MPCA method with the orthogonality constraint. Then, the semi-orthogonal MPCA results in more captured variance and more learned features than full orthogonality. In addition, this method can obtain correlation information among different color channels. In these experiments, the facial images in FERET database are used to test for a proposed method. The experimental results also indicate that the proposed method achieve better recognition rates than the well-known methods and it can be suitable for other color models such as HSV, YCbCr and CIELAB. Finally, the proposed recognition method can reduce the computational complexity in the color face recognition process.