Biomedical Engineering Research March. 2013, Vol. 2 Iss. 1, PP. 1-15 - 1 - A Region Growing Method Based on Statistical Attributes of Infrared Images for Finger Vein Pattern Extraction M. Vlachos *1 , E. Dermatas 2 Department of Electrical and Computer Engineering, University of Patras, Kato Kastritsi, 26500, Rio, Patras, Greece *1 mvlachos@teemail.gr; 2 dermatas@george.wcl2.ee.upatras.gr Abstract- In this paper a method for finger vein pattern extraction from infrared images of human finger is proposed. The finger vein pattern is extracted by the execution of a two-step region growing procedure, based on statistical properties of derivatives of the acquired infrared images. Initially, original image is filtered by four different Gaussian kernels (in order to take into account the different orientations of veins). Afterwards, the second partial derivatives of the obtained images are computed. Sequentially, the Hessian matrix of these images is constructed and its eigenvalues are computed in a pixel by pixel basis. The minimum eigenvalue and the absolute value of its gradient comprise the two characteristics (features) used in the two step region growing procedure which follows. The region growing procedure is restricted by statistical attributes such as the mean value and the standard deviation of the segmented regions (vein and tissue) and the mean value and the standard deviation of the gradient of the minimum eigenvalue image. Due to the occurrence of some misclassifications a final post processing step, based on morphological operations, is performed. The developed method achieves to efficiently segment the image despite of intensity variations which are evident in the original image. Moreover, an improved version of the proposed method, which uses the multidirectional response of a specially designed matched filter and its gradient as the two features used in the two stage region growing procedure, is also presented. The modified version, as experimental results show, outperforms the classic version and leads to more robust finger vein pattern extraction. Keywords- Vein Pattern; Region Growing; Hessian Matrix; Eigenvalue; Gradient; Matched Filter; Morphological Postprocessing I. INTRODUCTION The problem of finger vein extraction arises mainly for biometrics purposes but it is also very important for the biomedical research community. A low number of studies have been presented in the literature due to the small time distance from the first corresponding work. In the pioneering work of Park et al. [1] , an application specific processor for vein pattern extraction and its application to a biometric identification system is proposed. The conventional vein-pattern-recognition algorithm consists of a preprocessing part, applying sequentially an iterative Gaussian low-pass, a high-pass, and a modified median filter and a recognition part which includes the extraction of the binary veins via local thresholding and finally the matching between the individual patterns. Consequently the conventional algorithm [1, 2, 4] consists of low pass spatial filtering for noise removal, high pass spatial filtering for emphasizing vascular patterns, thresholding and matching. An improved vein pattern extracting algorithm is proposed in [3], which compensates 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. Also, the problem arising from the iterative nature of filtering preprocess is solved by designing a filter that is processed only ones, increasing significantly the recognition speed and reducing the hardware complexity. The proposed algorithm is implemented with a FPGA device and the Fa1se Acceptance Rate shows five times better than the existing algorithm and the recognition speed is measured to be 100 [ms/person]. The problem with conventional hand vascular technology mentioned above is that the vascular pattern is extracted without taking into account its direction. So, there is a loss of vascular connectivity which leads to a degradation of the performance of the verification procedure. An attempt to improve this problem can be found in [5], where 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 vascular pattern extraction, and column vascular pattern extraction 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 extraction approach prevents loss of the vascular pattern connectivity. Although, the above algorithm considers the directionality of veins, assumes that the veins oriented in only two principal directions. In [6-7] a method for personal identification based on finger-vein patterns is presented and evaluated using line tracking starting from various positions. This method allows vein patterns to have an arbitrary direction. 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 multiple times are classified as veins. The performance of the above method is strongly related to the number of repetitions. To achieve meaningful results, the