Color face recognition using adaptive feature fusion based on chroma mahalanobis distance map and gray map

The color provides much more information than the gray. Considering that the color facial image contains much more identification information than the gray facial image, the chroma Mahalanobis distance map was introduced into color face recognition. Based on the YCbCr color space, the information of chroma and brightness for one color facial image can be separated. Then Mahalanobis distance maps based on chroma information were constructed, at the same time, gray maps based on brightness information can be extracted by orignal color facial images. An algorithm for color face recognition using adaptive feature fusion based on chroma Mahalanobis distance map and gray map was presented. Based on energy values of wavelet packet sub-nodes, normalized feature vectors of chroma Mahalanobis distance maps and gray maps were constructed, respectively. Then identification feature vectors were constructed using several feature fusion methods, and fusion parameters were selected adaptively according to fusion effects. So feature complementation of chroma and brightness was achieved. Results of face recognition were obtained by the classifier based on the variance similarity degree. Experiments show that the algorithm has the characteristics of high recognition rate and good robustness.