Research Article Finger Vein Segmentation from Infrared Images Based on a Modified Separable Mumford Shah Model and Local Entropy Thresholding Marios Vlachos and Evangelos Dermatas Department of Electrical & Computer Engineering, Polytechnic Faculty, University of Patras, Rio Campus, 26504 Patras, Greece Correspondence should be addressed to Marios Vlachos; mvlachos@george.wcl2.ee.upatras.gr Received 3 September 2014; Accepted 26 January 2015 Academic Editor: Yi Gao Copyright © 2015 M. Vlachos and E. Dermatas. Tis is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. A novel method for fnger vein pattern extraction from infrared images is presented. Tis method involves four steps: preprocessing which performs local normalization of the image intensity, image enhancement, image segmentation, and fnally postprocessing for image cleaning. In the image enhancement step, an image which will be both smooth and similar to the original is sought. Te enhanced image is obtained by minimizing the objective function of a modifed separable Mumford Shah Model. Since, this minimization procedure is computationally intensive for large images, a local application of the Mumford Shah Model in small window neighborhoods is proposed. Te fnger veins are located in concave nonsmooth regions and, so, in order to distinct them from the other tissue parts, all the diferences between the smooth neighborhoods, obtained by the local application of the model, and the corresponding windows of the original image are added. Afer that, veins in the enhanced image have been sufciently emphasized. Tus, afer image enhancement, an accurate segmentation can be obtained readily by a local entropy thresholding method. Finally, the resulted binary image may sufer from some misclassifcations and, so, a postprocessing step is performed in order to extract a robust fnger vein pattern. 1. Introduction Te problem of fnger vein extraction from infrared images arises mainly for biometrics purposes but it is also very important for the biomedical research community. Te general structure of a biometric system based on fnger veins consists of fve main stages: (1) acquisition of the infrared images exploiting the absorption of light in near infrared and infrared wavelengths by the diferent human tis- sues, (2) preprocessing of the acquired images which includes ROI (region of interest) extraction, image intensity normal- ization (in this type of images intensity is usually uneven due to the image acquisition system and may sufer from shading artefacts), and noise reduction, (3) segmentation or classifcation stage in which the preprocessed image divided into two (or more depending on the application) regions asso- ciated with veins and surrounding tissues, (4) postprocessing of the binary images which delivers the fnal segmentation result, free of outliers and misclassifcations, and fnally (5) matching of the extracted veins in order to perform the desired person identifcation/verifcation procedure. Match- ing procedure can be applied either directly in the extracted fnger vein patterns or in their skeletons, depending on the matching algorithm that has to be used. Tis general structure described so far involves all the stages that may have such a system but it is worth mentioning that these stages are independent and some of them can be skipped in some applications depending on its specifc requirements. Related Work. Several methods which adopt this general architecture have already been presented starting from the pioneering work of Park et al. [1]. In this important research work, an application specifc processor for vein pattern extraction and its application to a biometric identifcation system is proposed. Te conventional vein pattern recogni- tion algorithm [1–3] consists of a preprocessing part, applying Hindawi Publishing Corporation Computational and Mathematical Methods in Medicine Volume 2015, Article ID 868493, 20 pages http://dx.doi.org/10.1155/2015/868493