Abstract: A new fingerprint compression technique based on sparse representation is introduced. Obtaining an over complete dictionary from a set of different fingerprint patches allows us to represent them as a sparse linear combination of the dictionary atoms. In the algorithm, first step is to construct a dictionary for predefined fingerprint image patches. For a new fingerprint images, represent its patches according to the dictionary by evaluating l0-minimization problem and then quantize and encode the matrix. In this paper, we consider various factors affecting the compression results. The main idea behind the project is to construct a base matrix whose columns represent features of the fingerprint images, referring to the matrix dictionary whose columns are known as an atoms; for a given whole fingerprint image, create a patches whose number of pixels are equal to the dimension of the atoms; use sparse representation method to obtain the coefficients; then, quantize the coefficients; and last, encode the coefficients with the help of lossless coding methods. Especially at high compression ratios the proposed algorithm is efficient compared with several competing compression techniques like (JPEG, JPEG 2000, and WSQ).
Keywords: Fingerprint, compression, sparse representation,PSNR JPEG 2000, JPEG, WSQ.