The face recognition is nowadays popular in social networks and smart phones. The face recognition is more difficult for poor illumination images. The objective of the work is to create an illumination invariant face recognition system using 2D Discrete Cosine Transform and Contrast Limited Adaptive Histogram Equalization (CLAHE). Contrast Limited Adaptive Histogram Equalization is used for enhancing the poor contrast medical images. The proposed method selects 75% to 100% DCT coefficients and set the high frequency to zero. It resizes the image based on the selection percentage, and then inversed DCT is applied. Then, CLAHE is applied toadjust the contrast. The resized images reduce the computational complexity. The image obtained is illumination invariant face image and termed as ‘En-DCT’ image. The fisher face subspace method is applied on the ‘En-DCT’ image to extract the features. The matching face image is obtained using cosine similarity. The face recognition accuracy is tested on ARdatabase. The face recognition is tested with 75% to 100% DCT coefficients and finds the bestrange. The performance measures recognition rate, 1% FAR (False Acceptance Rate) and Equal Error Rate (EER) are computed. The high recognition rate results prove that the proposed method is an efficient method for illumination invariant face recognition.
展开▼
