Point Based Fea Anil Singh Parihar 1 , Amoiy Kuma 1 Department of Informatio 2 Department of Elect parihar.anil@gmail.com, amio mukherjee Abstract— In this paper, feature extraction an scenarios for contact-based and contact-less p are investigated. The point-based feature extr like: Scale Invariant Feature Transform (SIFT detector, and Histogram of Gradient (HOG) i Gabor filter are experimented for contact-based palmprint authentication. In our experiment publicly available IITD database consisting contact-less palmprint images and compared its these features with HongKong PolyU database constant illumination and constrained condition work establishes that majority of the previous m print authentication, may work well with conta (with constrained environment) but fail to pro results with unconstrained natural contact-less p The proposed techniques give high recognition 93% GAR with Harris and SIFT respectively palmprint database. The experimental results in shown significant improvement for contact-ba well. Key words: palmprint, contact-based, contact- ROI I. INTRODUCTION Biometric authentication refers to the automa of a person based on human physiological ( iris) or behavioural (e.g., signature) or chemic composition of human perspiration) charact [34]. The traditional methods require the pin password to be remembered for identificatio and thus are more vulnerable to attacks. As a systems are being deployed to enhance secu financial fraud [34][43]. Various biometric used for real-time recognition, the most fingerprint, face and iris [41][44]. Amo biometric techniques used for authenticati biometrics is well established trait, with some the already prevalent competitor techniques. of the textural details which can be extracted features for authentication is high and the h acquisition is convenient and user-friendly. exposed to anatomic variations and environ Higher recognition rates in most hand based m a contact device with pegs for image acqu based biometrics has been used in personal using fingerprint[44],palmprint [5][29-33], atures for Contact-less P Images ar 2 , Om Prakash Verma 1 , Ankita Gupta 1 , Prerana Mukherj on Technology, Delhi Technological University, New Del trical Engineering, Indian Institute of Technology, New D oy.iitd@gmail.com, opverma.dce@gmail.com, ankita.gup e.prerana@gmail.com, vatsa.deepika@gmail.com nd authentication palmprint images action techniques T), Harris corner n combination to d and contact-less ts, we have used of unconstrained s performance on e acquired under ns. The presented methods for palm act-based images, oduce substantial palmprint images. n rate (97.5% and y) for contact-less n this paper have ased palmprint as -less, constrained, N atic identification (e.g., fingerprint, cal (e.g. chemical teristics or traits n number or the on/authentication result, biometric urity and reduce traits are being t popular being ong the various ion, hand based e advantages over It is so, because from hand based human hand data Also, it is less nmental artifacts. modalities rely on uisition[5].Hand- identification by hand geometry [15][35], 3-D finger geometry [36- finger knuckles [6-9]. Fingerprint is modality since it is easy to use a However, as investigated by NIST population do not have useful and fingerprint identification is usele Palmprint is a highly accurate biom larger surface area for feature extra more information for personal capture devices are much low in cos is less for the storage of low resolut The literature works show th efforts are on contact-based palm growing concern over using co palmprint images as it does not gi scenario. It has resulted in inclin towards constrained free natura authentication. The contacts-les systems are more convenient to th hygiene issues. However, a contact may suffer from scale, rotation, o variations which can reduce auth available approaches for palmpri categorized into following cate approaches (Gabor filter, Discrete 22] (ii) line-based approaches (Lin Operators) [23-24], and approaches(PCA, LDA)[25-27]. M have shown satisfactory results on difficulties faced by user in contac due to constrained applied like app in orientations, and necessity of c may be unhygienic. Nowadays, a idea of peg-free, contact-less, hand The contact-less imaging increas [30][37][39], enhances real life sc imposter attacks. There are various contact-less system. It involves a c between the hand and input senso rotational variations, translational v blurring etc., since there is no rest during image acquisition [42].T traditional palmprint feature extrac less imaging schemes remains ques almprint jee 1 , and Deepika Vatsa 1 lhi, India Delhi, India pta.mail@gmail.com, 37], hand vein [38-40] and s a widely used and reliable and gives accurate results. T, approximately 2% of the accessible fingerprints thus ess for such people [1]. metric modality as there is a action, and thus they carry identification. Moreover, st and memory requirement ion palm images. hat, most of the available mprint acquisition. There is ontact-based (constrained) ve a user friendly real life nation of research interest al contact-less palmprint ss palmprint acquisition he user and have negligible t-less palmprint acquisition occlusion and translational hentication accuracy. The int authentication can be egories: (i) texture-based Fourier Transform) [5][21- e matching, Morphological (iii) appearance-based Most of the reported works n constrained images. The ct-based system, are mainly plication of pegs, limitation ontact with surface, which growing trend towards the d biometrics has emerged. es user convenience [28] cenario and results in less s challenges in designing a areful selection of distance or and adds variations like variations, scale variations, triction on hand placement Therefore, the usage of ction methods on contact- stionable and hence most of 978-1-4799-1535-4/13/$31.00 ©2013 IEEE 165