Fingerprint Detection Applying Discrete Wavelet Transform on ROI

iometric identification techniques are applied to identify an individual on the basis of an individual’s characteristics (both physiological and behavioral). It is one of the most dependable and sensible approach to recognize an authorized person among several masquerades. The technology of biometric identification is applying in some specific regions to draw out furtive information such as face and eye structure, handwriting, signature, voice, hand and finger geometry, fingerprint as well as palm-print imaging [1]. The process of identifying the human fingerprint where ridge skin layout (minutiae) is used is also known as dactyloscopy. Specifically, ridge and furrow patterns on the surface/tip of the finger including bifurcations, termination and valley have been applied in a comprehensive way to determine the uniqueness of fingerprint of human being. Bifurcations are identified by the branch of one ridge, termination is the endpoint of ridge and valley is the gap between two ridges. Small and precise portions called minutiae represents those local ridge characteristics, which is also called singular points or singularities. This minutiae-based approach has minimized the tiresome job of manual classification and matching method. Therefore, according to the methodology of Henry Classification System [2], [3], there exist three main fingerprint textures: loop, whorl and arch. Generally, Automatic Fingerprint Identification Systems (AFIS) performs three basic steps to recognize fingerprint: pre-processing, region of interest (ROI) extraction and finally classification [4], [5]. Several approaches of fingerprint matching have been proposed in recent literatures. One approach is taking the 9th level DWT of the original fingerprint image; where the slopes of the three linear lines are obtained and stored as matrix form and used as template value for comparing with others explained in [6]. The fingerprint images are matched based on the features extracted in the wavelet domain. Another, hierarchical fingerprint matching system is proposed in [7] that utilized features at three levelsLevel 1(pattern), Level 2(minutia points) and Level 3 (pores and ridge contours), extracted from high resolution fingerprint scans. Here Gabor filters and wavelet transform are used to automatically extract the Level 3 features and are locally matched using Iterative Closest Point (ICP) algorithm. K. Thaiyalnayaki et al [8] propose a combination of features (standard deviation, kurtosis, and skewness) for multi-scale and multi-directional recognition of fingerprint, uses Canberra distance metric to determine the similarity between the texture classes. Fingerprint identification using Wavelet Transform (WT) and Probabilistic Neural Network (PNN) is proposed in [4] where the feature vector is obtained by performing onedimensional DWT. Gabor Filter based fingerprint classification using support vector machines (SVM) is proposed in [2]. A one step method using Gabor filters for directly extracting fingerprint features from grey level images for a small scale fingerprint recognition system is introduced by C.J. Lee and S.D. Wang [9]. In [10], an identification system uses a gray level watershed method based on edge detection to find out the ridges present on a particular fingerprint image for compare. B. Tan et al [1] present a detection method based on noise analysis along the valleys in the ridge-valley structure of fingerprint images. A neural network based method is applied and trained in [11] throughout the fingerprint skeleton to locate various minutiae. Another fingerprint recognition based on wavelet domain features is presented in [12]. The features are directly extracted from the wavelet transform and the recognition performance achieved by using the proposed wavelet features were evaluated using a K-NN classifier. In this paper we used combination of ROI, binarization of an image and DWT to extract the feature of a fingerprint. The paper is organized like: section 2 deals with theoretical approach of discrete wavelet transform (DWT) and its application in feature detection of fingerprint is shown by a flowchart with 8 steps, section 3 gives the results of the paper and finally section 4 concludes the entire analysis. B