Fingprint Identfication Based On Wavelet Transform

As the rapid development of science and technology, how to authenticate identity accurately effectively and timely has become a hot topic. As a reliable biometric technology, fingerprint identification technology gets more and more attention. Quality of fingerprint image acquired is often poor, because a finger maybe dry,wet, burglary skin, scars, knife-edge and so on. Performance of fingerprint identification system is directlt impact by this matter, while the fingerprint image recognition algorithm is also put forward to higher accuracy.In this paper, in view of characteristics of fingerprint images, wavelet transform, neural network algorithm, genetic algorithm are used in fingerprint identification technology, to improve the performance of fingerprint recognition algorithm. The main contributions of this dissertation are summarized as follows:1 In view of the drawback of the local grayscale variance algorithm, point ratio function is exploited to serve as the parameters for image segmentation, which are calculated based on the thumbprint area that is extracted using the multi-scale analysis. Furthermore, the thumbprint district is processed by overrun threshold function to separate ridge from valley and reduce the fuzzy distortion of the fingerprint image. Experimental results are provided to show that the proposed algorithm is anti-noise, robust, and outperforms the conventional local grayscale algorithm.2 In view of the information of direction of fingerprint images a fingerprint image enhancement algorithm based on wavelet transform and median filter was proposed. Variance minimum criteria and principle of directional spread are used to extract characteristic data from fingerprint image, enhancing relevance of the information. According to noise statistical nature in the wavelet transform, methods of median filter and atrophic wavelet soft threshold are used for characteristic data de-noising. The simulation results demonstrate that by this algorithm fingerprint image is de-noised effectively, ridge is enhanced obviously.3 A fingerprint image compression algorithm based on wavelet and neural network encoded is proposed. After wavelet transform the high-frequency sub-band of fingerprint image is divided into scan blocks, the scan block is divided into categories of weak transition smooth texture and strong texture, using neural network methods. Then important wavelet coefficients are scanned based on image texture complexity and the importance of the extent. Important wavelet coefficients are coded using embedded zerotree wavelet (EZW) coding. Experimental results show that the image compression algorithm is superior to DCTand JPEG encoding algorithm in image compression rate and recovery quality.4 Optimal corresponding relation between input minutiae and template minutiae is searched using genetic algorithm. The number of corresponding minutiae points between two minutiae points modes is largest by this corresponding relation, while matching error of corresponding point is smallest. Match score is calculation according to the number of corresponding minutiae points, matching error and a series of logical rules.