Abstract:
Face-Recognition is one of the most efficient and highly popular technology in biometrics
because of its least intrusiveness, reliability and easy use. We introduce a novel algorithm
namely, Robust NCC, for face recognition with varying illumination, expression, occlusion,
disguise.
RNCC has exceptionally remarkable results. It‟s fast and can correctly identify highly
occluded images without adding much complexity.
To deal with illumination variation, the accuracy rate is further improved by cascading it with
Collaborative Representation Classifier (CRC). In order to use the two classifiers in fusion
setting, we perform intelligent cascading through confidence weighted scheme.
The final cascaded method is tested on 7 renowned databases (AR, Extended Yale B, Cohn
Kanade and Cohn Kanade plus, Bosphorus, Yale Faces, Jaffe). It outperforms state of the art
Sparse Representation and other well-known classifiers. The recognition rate for all the tested
databases with low dimensionality (13 x 10) is above 90%.
