T R ANOOP AND M G MINI: TWO STAGE FRAMEWORK FOR ALTERED FINGERPRINT MATCHING 1024 TWO STAGE FRAMEWORK FOR ALTERED FINGERPRINT MATCHING T.R. Anoop 1 and M.G. Mini 2 Department of Electronics and Communication Engineering, Model Engineering College, India E-mail: 1 anooptr234@gmail.com, 2 mininair@mec.ac.in Abstract Fingerprint alteration is the process of masking one’s identity from personal identification systems especially in boarder control security systems. Failure of matching the altered fingerprint of the criminals against the watch list of fingerprints can help them to break the security system. This fact leads to the need of a method for altered fingerprint matching. This paper presents a two stage method for altered fingerprint matching. In first stage, approximated global ridge orientation field of altered fingerprint is matched against the orientation field of its unaltered mate. If this matching is successful, fingerprints go to second stage. Second stage starts with the selection of unaltered region from the altered FP and same region from the unaltered mates. Matching in this stage is performed by extraction of ridge texture and ridge frequency from the selected region of interest. Euclidian distance is used in both stages to compute the matching score. Keywords: Matching, Ridge Orientation, Ridge Texture, Ridge Frequency, Wavelet Transform 1. INTRODUCTION Fingerprint alteration is a major threat to boarder control security systems [1]. Altered fingerprint (FP) is entirely different from fake fingerprints. Fake fingerprints are made from latex or glue for imitating the normal fingerprints so that the matching become successful and they can break the security system. Different ways of making altered fingerprints are abrasion and cutting with blades, poring chemicals and transplantation of ridge structure by surgery. Based on these, altered fingerprints are classified into three types [1], [2], obliteration, distortion and imitation. Processes like abrasion and cutting with blades and poring chemicals leads to obliteration. It is again divided into scar and mutilation. Even though the distortion and imitation are produced by surgery, they are different in terms of area of transplantation. Imitation contains large area transplanted from palm print or leg print. Distortion contains small area transplanted from different portion of the same fingerprints. Fingerprint recognition is basically divided into three types [3]. They are correlation based, minutiae based and texture features based techniques [3]. In correlation based techniques, two FP images are superimposed and the correlation between corresponding pixels is computed for different alignments [4-8]. Ridge endings and bifurcation is known as minutiae features and is unique for each FP. Minutiae-based matching consists of finding the alignment between the template and the input minutiae feature sets that result in the maximum number of minutiae pairings [9-14]. In non minutiae based matching, features of the ridge pattern like local orientation, frequency, ridge shape and texture information and number, type and position of singularities are used [15-17]. Successful matching of altered fingerprint (FP) is essential since it helps to find the criminals and also prevents the breaking of fingerprint based security system. Altered FP matching can be done in two ways. First method needs the reconstruction or restoration of altered region and matching is performed in the same way as the normal FP matching. In second method, matching can be done by using the features in the unaltered region. These two methods fails when whole region of the fingerprint is altered. Proposed method is a hybrid one since it uses the reconstructed ridge orientation and features in the unaltered region. Matching of imitation type altered FP is not possible since reconstruction of altered region is not possible. Separation of unaltered region from the altered region is also difficult. Matching of Z-type distortion is possible because altered region can be reconstructed. Soweon Yoon and Anil K. Jain have performed the matching of „Z‟ cut type of distortion by the restoration of minutiae distribution in the altered region [18]. Proposed method belongs to texture feature based technique. Rest of the paper is organized as follows. Proposed method starts from section 2. Section 3 explains approximation of ridge orientation field of altered FP and alignment. Matching score computation using orientation field is given in section 4. Region of Interest extraction is given in section 5. Section 6 explains the ridge frequency extraction. Section 7 explains matching score computation from ridge frequency and ridge texture images. Results and discussion is given in section 8 and conclusion is given in section 9. 2. PROPOSED METHOD The method uses three features namely Ridge Orientation Field (ROF), Ridge Texture (RT) and Ridge Frequency (RF). Last two features are extracted from the unaltered region of the altered FP. The computation of a single matching score from these three features is not possible since automatic selection of unaltered region from the altered FP is not possible in one to many matching. FP images from different fingers may sometimes appear similar in terms of global structure such as local ridge orientation [3]. Thus the proposed method is implemented in two stages. First stage utilizes the approximated ridge orientation to compute the matching score in terms of Euclidian distance and if this matching is successful, the FP goes to next stage. Second stage is used to confirm the successful matching of the first stage by using RT and FR in the unaltered region. A matching is declared as successful, if genuine match occurs in both the stage. Fig.1 shows the block diagram of the proposed method.