Fingerprint Compression using Singular Value Decomposition

Kiran B. Khedkar, Prof. A. S. Deshpande

Abstract: A Fingerprint compression technique is becoming more and more for security purpose required less space to store the data so I am processing new fingerprint compression technique using sparse representation obtains the complete dictionary from the set of fingerprint patches allows to represent sparse linear combination of dictionary atom. In the algorithm, we first collect the fingerprint patches and construct a dictionary for predefined fingerprint image patches It uses the l0-mimimization technique to represent the fingerprint patches according to the dictionary and then quantize and encode the representation. Those having the Coefficient less than given threshold treated as Zero and record the remaining coefficient and their locations. Then encode the atom number of each patch and find out the mean value of each patch. In this we also compare our method with other techniques such as JPEG JPEG2000 and Wavelet Scalar Quantization especially at high compression ratios. The experiments also illustrate that the proposed algorithm is robust to extract minutiae.

Keywords: Fingerprint compression, sparse representation, Singular Value Decomposition, JPEG, JPEG2000, WSQ