An Improved Algorithm for Fingerprint Compression Based on Sparse Representation

Abstract

An improved algorithm to compress fingerprint images based on sparse representation is proposed. The algorithm includes two parts namely, construction of the dictionary and the compression process. In order to construct the dictionary, recursive least squares dictionary learning algorithm (RLS-DLA) is used. In RLS-DLA, any given fingerprint is divided into small blocks called patches. Then sparse coding is performed on each patch and the dictionary is continuously updated. Each patch is represented as a linear combination of a few columns from the pre-constructed fingerprint dictionary, which leads to compression. To compute a linear expansion of the current patch, orthogonal projection of the patch on the pre-constructed dictionary element is done. Then the representation is quantized and encoded. The results obtained through RLS-DLA shows improvement of 2.98% in PSNR compared to K-singular value decomposition (K-SVD) dictionary learning algorithm.