Facial recognition by Optimal Random Image Component Selection

An innovative approach based on local components called Optimal Random Image Component Selection is presented in this paper. Here, features are extracted from the Optimal Random Image Components by Gabor wavelets using greedy approach is proposed. These feature vectors are then down-sampled to some size which is then classified based on minimum distance measure. The design of Gabor filters for facial feature extraction is also discussed. The FERET face database is used to generate the results. Experiment shows that ORICS outperforms local features in terms of expression, pose, illumination and occlusion. Our method has achieved 100% recognition accuracy on the FERET database of the image of size 64 × 80. This is a considerably improved performance than that attainable with other standard methodologies described in the literature.