Research On Fingerprint Image Compression System

As one of the most commonly used biometrics technology, fingerprint recognition has been widely adapted in identification and verification of a person. As the number of the fingerprint templates has been increased, it will degrade the performance of the fingerprint verification system. One way to improve the efficiency of the system is to reduce the template size by image compression on the condition that the compression time is sufficiently small. For compression algorithms, the higher compression ratio and fidelity of these algorithms will incur the complexity of computation.Firstly, a compression methodology based on Wavelet Scalar Quantization (WSQ) algorithm for fingerprint compression algorithm has proposed which is using 5/3 Lifting Scheme Discrete Wavelet Transform (LS-DWT) in this case. This is actually a kind of image compression algorithm, including 3 level 5/3 LS-DWT, quantization, Huffman coding and run-length coding. The simulation results showed that: the proposed algorithm is better than the traditional WSQ with respect to compression ratio and image quality. e.g., when compression ratio is 16(0.5bpp) , the proposed algorithm can reach a Peak Signal-to-Noise Ratio (PSNR) of 57.34dB and has a lower computational complexity, thus making it much easier to in hardware implementation.Secondly, compared with those image compression methods, the minutiae extraction in fingerprint images can obtain a higher compression ratio because it just keeps the characteristic information instead of storing the whole original fingerprint image. However, computational complexity is higher and the process is tedious. The algorithm is successfully transplanted to C language, which allows of the algorithm to be implemented on Nios II soft-core processor platform.Finally, we have built partly the fingerprint scan and storage for the whole fingerprint image compression system on an EP2S60 FPGA (field programmable gate array) development board with the Quartus II 6.0 integrated development platform. The two subsystems are necessities for the further compression of the fingerprint images, which include the HV7131 fingerprint sensor. It is up to 1361 ALUTs, 1057 registers and 2 PLLs being used in the design. In addition, the fingerprint scan and storage subsystem can communicate with PCs via a Universal Asynchronous Receiver/Transmitter (UART).