One-sample face recognition with local similarity decisions

Abstract

Inter-class and intra-class variability account for poor classification performance of automatic image recognition methods under the condition of limited number of gallery images per class. Texture-based intensity features in combination with local decisions on similarity measurements are used to reduce the effects of variability and to provide a robust image recognition method. The presented method show good recognition performance when tested under the most difficult condition of using single gallery image per class on AR, YALE , EYALE and FERET 2D face image databases.