A novel approach to eigenpalm features using feature-partitioning framework

Abstract

Eigenpalms, a well-known approach, extracts features from palmprint images using conventional PCA technique. However eigenpalms does not exploit neighbourhood (local) information due to its vector representation of palmprint images. In our work here, we propose a feature-partitioning framework that uses a more efficient and appropriate matrix representation of images. Our novel feature partitioning approach shows a considerably better and consistent recognition performance than eigenpalms approach (PCA).