Research On Fingerprint Image Compression Algorithms And Its Applications

At present, the automatic fingerprint recognition is the most widely used and relatively maturetechnology among all the biometric recognition technologies. Usually, a typical automaticfingerprint identification system (AFIS) includes fingerprint image acquainting, preprocessing,feature extracting, matching etc. The original fingerprint images are collected by live fingerprintsensor. Therefore, with the fingerprint image samples increases continuously, these large fingerprintimage data will bring a huge burden to the system memory medium and the communicationbandwidth. Especially for some large-scale applications, the problem will be more outstanding.Such as the acquisition and storage work of the criminal fingerprint database in the public securitysystems, driver fingerprint database in traffic systems, and so on. Obviously, the fingerprint imagecompression can effectively solve this problem. In this paper, based on the analysis ofsome classical wavelet-based still image compression algorithms, we focus on the fingerprint imagecompression algorithms and the applications of wavelet-based compression algorithms infingerprint image spectrum recognition. The main attributions in this paper are as follows:(1) Analyze some classical wavelet-based still image compression algorithms, and introducethe components and core technologies of the latest international imagecompression standard JPEG2000.(2) Based upon the FBI’s WSQ algorithm, an improved fingerprint image compressionalgorithm IWSQ is proposed.The main points of the proposed algorithm are: 1) The simple lifting steps are employed in thewavelet decomposition of the fingerprint image; 2) The wavelet coefficients are processed by anefficient adaptive scalar quantization scheme based on the energy of the sub-bands; 3) The encodingmodules are divided by the characteristics of wavelet coefficients and the specificquantitative scheme, and the appropriate encoding mode is selected for each coding module; 4)According to the characteristics of the output Huffman stream in the high-frequency sub-bands, anefficient adaptive run-length encoding is used in the IWSQ algorithm. In order to maximize theimage compression ratio, the specific parameters of the coding algorithm are determined by thesimulation results.The implementation of the proposed fingerprint compression algorithm IWSQ is completedwith C ++ language in the platform VS2005. We compare the performance of our algorithm IWSQto WSQ and EZW in the platform MatlabR2001a. The results show that the proposedalgorithm’s PSNR improves 2~4dB compared to the WSQ algorithm at the same bit rate. Through minutiae extraction for the reconstructed fingerprint images, the results prove that theIWSQ algorithm can more effectively protects the fingerprint minutiae compare to EZW algorithm.(3) The POC algorithm and the BLPOC algorithm for the fingerprint image spectrumrecognition are analyzed. And based on the BLPOC algorithm, the FMT is used to correct therotation of the fingerprint image in our spectrum recognition scheme. In order to show the goodperformance of the spectrum recognition scheme to low-quality fingerprint images, acommercial fingerprint recognition algorithm which is based on the pattern of minutiae is used tocompare with it. At the same time, to solve the storage problem of the fingerprint templates in thespectrum recognition system, we focus on the applications of the wavelet-based compressionalgorithms in fingerprint image spectrum recognition.(4) We summarize the contents of the paper and point out the problems in the study and ourfuture work.