Face recognition using Shearlets

In this paper we propose a novel statistical face recognition method that uses a new multiresolution analysis called Shearlet transform for facial texture features representation. In recent years Shearlet transform has emerged as the most successful framework for the efficient representation of multidimensional data in which directional information is exploited along with the conventional scaling and translation parameters as in wavelets. Features are computed by low order statistics like mean and covariance of transformed face images. Then, an efficient and reliable probabilistic metric derived from the Bhattacharyya distance is used to classify the extracted feature vectors into face classes. The efficiency of the algorithm is tested on ORL database. Efficiency of the proposed approach is demonstrated with exhaustive experiments.