Fingerprint Image Enhancement Algorithm Research And Accelerator Design

With the increasing requirement of information security, Biometric Identification Technology will replace the traditional security measure of ID plus password, and give its popularity. Due to its unique properties and easy acquirement, Fingerprint Identification Technology work as the main part of Biometric Identification Technology. As the core technology of commercial fingerprint identification algorithms, The technology of fingerprint image enhancement must be effective as well as efficient. In this paper, we describe the general enhancement algorithms, sum up their benefit and drawback, and give a discussion on how to choose parameters. Because of its perfect performance both on special domain and frequency domain and insignificant block effect, Gabor filter is one of the most popular filters used to enhance the fingerprint image. However, it consumes a great deal of computation, which becomes a bottleneck in embedded system. In order to satisfy the application of on-line fingerprint identification system, we modified the algorithm, accelerate it by hardware implementation, and evaluate its performance. According to the experiment, our solution can meet the need of real-time embedded system, and improved the power consumption by optimizing the data path and data access. Additionally, this algorithm can be used in the optimization of software.To improve the quality of image enhancement, based on the traditional algorithm, we propose an improved Gabor filter and a new method for frequency estimation with DCT. Furthermore, we extend the use of DCT, and give a new algorithm on image partition and singular points location.The experiments have indicate that, our algorithm gets more qualified image, more accurate frequency estimation and more effective partition than the traditional one, and can locate the singular points with little computing cost.