COMPARATIVE ANALYSIS OF FEATURE EXTRACTION METHODS BASED ON DWT, DTCWT AND HOG USED FOR FINGERPRINT RECOGNITION

Fingerprint Recognition is one of the most effective and common biometric methods used to identify individuals. In forensic and civil applications such like criminal identification, access control and ATM card verification it plays a very important role. Image-based and minutiae-based are the two main methods used for fingerprint recognition. The commonly used comprehensive representation does not allow full use, however, of a substantial component of the rich, unequal knowledge available in fingerprints and local ridge structures. Furthermore, careful matching has difficulty in rapidly matching two fingerprint images with various unregistered point numbers. This paper details a comparative fingerprint recognition analysis using DWT, Dual Tree Complex Wavelet Transformation (DTCWT) and Histogram of Gradient (HOG) feature extraction methods. The image-based algorithms are used as a compact fixedlength feature to capture both local and global details as fingerprints to overcome the inconveniences of minutiae based fingerprinting identification. The Distance of Euclidean (ED) is used to compare test image with data base image features. The performance of the recognition is measured with respect to false acceptance rate (FAR), false recognition rate (FRR) and equal error rate (EER). The performance measurements show that the HOG-based method results better than the other two