FINGERPRINT FEATURE EXTRACTION BASED DISCRETE COSINE TRANSFORMATION (DCT)

INTRODUCTION

Among various biometrics technologies, automatics fingerprint identification systems are the most popular and reliable for automatics personal identification system . Fingerprint identification system has received increasing attention and the performance of fingerprint identification system has achieved a high level since year of 2000 due to the need of security in a large range of application . Many automatics fingerprint image recognition and classification methods are already available in the literature. Even many academic and commercial systems for fingerprint recognition exits, there is a necessity for further research in the topic in order to improve the reliability and performances of the fingerprint recognition system. Most of the fingerprint recognition system is based on the minutiae and feature extraction. The common extracted minutiae are ridge ending and bifurcation minutiae. This is because these two types of minutiae are found on most of the fingerprints. Generally, the fingerprint recognition systems are based on the matching of extracted ridge ending and bifurcation minutiae. The extracted minutiae provided the information of position, types and quantities. Most of the matching process is done on the matching of the position and the total number of detected minutiae. Generally, the automatics fingerprint identification systems required the input fingerprint to be match with a large number of fingerprints in the database.

In additional , the fingerprint recognition systems are unable to solve the problem of rotated input fingerprint image. The matching systems are failed to match the rotated input fingerprint image with the fingerprint image that stored in the database, which both of the fingerprint images are actually belonging to the same person. An effectiveness and efficient technique is required in order to solve the problem. In this project, a simple feature extraction technique based DCT is proposed and the proposed technique is able to solve the problem of  matching processes based rotated input fingerprint images. Further, the extracted DCT data are used as input for neural network training and this technique is able to reduce the computation and searching time for automatics fingerprint recognition.