P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 741–744, 2010. www.springerlink.com Supervised and Unsupervised Finger Vein Segmentation in Infrared Images Using KNN and NNCA Clustering Algorithms M. Vlachos and E. Dermatas Department of Electrical Engineering and Computer Technology, University of Patras, Patras, Greece Abstract— In this paper, two new methods to segment infrared images of finger in order to perform the finger vein pattern ex- traction task are presented. In the first, the widespread known and used K nearest neighbor (KNN) classifier, which is a very effective supervised method for clustering data sets, is used. In the second, a novel clustering algorithm named nearest neighbor clustering algorithm (NNCA), which is unsupervised and has been recently proposed for retinal vessel segmentation, is used. As feature vectors for the classification process in both cases two features are used: the multidirectional response of a matched filter and the minimum eigenvalue of the Hessian matrix. The response of the multidirectional filter is essential for robust classi- fication because offers a distinction between vein-like and edge- like structures while Hessian based approaches cannot offer this. The two algorithms, as the experimental results show, perform well with the NNCA has the advantage that is unsupervised and thus can be used for full automatic finger vein pattern extraction. It is also worth to note that the proposed vector composed only of two features is the simplest feature set which has proposed in the literature until now and results in a performance comparable with others that use a vector with much larger size (31 features). NNCA also quantitatively evaluated on a database which con- tains artificial images of finger and achieved the segmentation rates: 0.88 sensitivity, 0.80 specificity and 0.82 accuracy. Keywords— KNN, NNCA, Vein pattern, Hessian matrix, matched filter, morphological postprocessing. I. INTRODUCTION An application specific processor for vein pattern extrac- tion, and its application to a biometric identification system, is proposed in [1]. The conventional vein-pattern-recognition algorithm consists of an original image grab part, a preproc- essing part, and a recognition part, and the last two parts take most of the processing time. The preprocessing part consists of a Gaussian low-pass filter (which works iteratively), a high-pass filter, and a modified median filter. Consequently the conventional algorithm [1, 2, and 4] consists of low pass spatial filtering for noise removal, high pass spatial filtering for emphasizing vascular patterns and thresholding. An im- proved vein pattern extracting algorithm is proposed in [3] which compensate the loss of vein patterns in the edge area, gives more enhanced and stabilized vein pattern information, and shows better performance than the existing algorithm. In [5], a direction-based vascular pattern extraction algorithm based on the directional information of vascular patterns is presented for biometric applications. It applies two different filters: row vascular pattern extraction filter for abscissa vas- cular pattern extraction, and column vascular pattern extrac- tion filter for effective extraction of the ordinate vascular patterns. The combined output of both filters produces the final hand vascular patterns. Unlike the conventional hand vascular pattern extraction algorithm, the directional extrac- tion approach prevents loss of the vascular pattern connec- tivity. In [6-7] a method for personal identification based on finger-vein patterns is presented and evaluated using line tracking starting from various positions. Local dark lines are identified and line tracking is executed by moving along the lines pixel by pixel. When a dark line is not detectable, a new tracking operation starts at another position. This procedure executes repeatedly, so the dark lines that tracked a lot of times have a great probability to be veins. An algorithm for finger vein pattern extraction in infrared images is proposed in [8]. The low contrast images, due to the light scattering effect, are enhanced and the fingerprint lines are removed using 2D discrete wavelet filtering. Kernel filtering produces multiple images by rotating the kernel in six different direc- tions, focus into the expected directions of the vein patterns. The maximum of all images is transformed into a binary image. Further improvement is achieved by a two-level mor- phological process: a majority filter smoothes the contours and removes some of the misclassified isolated pixels, and a reconstruction procedure removes the remaining misclassi- fied regions. The final image has segmented into two regions, the vein and the tissue. In [9] a certification system that com- pares vein images for low-cost, high speed and high precision certification is proposed. The equipment for authentication consists of a near infrared light source and a monochrome CCD to produce contrast enhanced images of the subcutane- ous veins. As recognition algorithm used only phase correla- tion and template matching. In [10], the theoretical founda- tion and difficulties of hand vein recognition are introduced at first. Then, the threshold segmentation method and thinning method of hand vein image are deeply studied and a new threshold segmentation method and an improved conditional thinning method are proposed. An initial work for localizing surface veins via near-infrared (NIR) imaging and structured light ranging is presented in [11]. The eventual goal of the system is to serve as the guidance for a fully automatic (i.e., robotic) catheterization device. The proposed system is based