Affine normalized stockwell transform based face recognition

In this paper, we have developed a new local descriptor based on Stockwell transform. The descriptor is made scale and affine normalized so that it is suitable for different types of applications. An experimental investigation has been conducted considering face recognition problem. The sparse representation based classification scheme is used for face classification. By extracting the facial features in the form of scale and affine normalized local regions around key points and representing those regions in terms of the Stockwell transform along with sparse representation based classification has shown promising results for face recognition. Experimental results conducted on the FERET and ORL datasets have demonstrated the suitability of the proposed method for face recognition. Comparative analysis is also made with the recently developed face recognition approaches.