Image compression based on SVD for BoVW model in fingerprint classification

Abstract

Fingerprint is the oldest techniques for biometric identification. This limb has its own characteristics for each individual to distinguish from each other. Fingerprint is unique and the pattern will never change. Many approaches have been developed in this field research. Bag-of-Visual Words (BoVW) is an approach for fingerprint classification, where the classification is based on the histogram of the frequency of visual words. The problem arises when the fingerprint classification uses the original images in which these images have a large size and in large numbers. It requires large storage space and high computational time. In order to overcome this problem, we propose Singular Value Decomposition (SVD) as compression method that can reduce the storage space. Our proposed method has been evaluated with some experiments based on five different types of fingerprint (whorl, arch, left loop, right loop, and twin loop). Experimental results show that the proposed method reduces the storage space and computational time. Moreover the accuracy of fingerprint classification could be improved.