Egyptian Computer Science Journal Vol. 39 No. 2 May 2015 ISSN-1110-2586 -72- An Approach for Pit Pattern Recognition in Colonoscopy Images Veska Georgieva 1 , Szilvia Nagy 2 , Andras Horvath 3 , Elena Kamenova 4 1,4 Technical University of Sofia, Bulgaria, 2,3 SzechenyiIstvanUniversity of Györ, Hungary vesg@tu-sofia.bg, nagysz@rs1.sze.hu,horvatha@sze.hu,elena.kamenova@best.eu.org Abstract In this worka new multistage approach is proposed for pit pattern recognition and classification of colonicmucosa, based on Kudoetal. Classifications system. Itconsistsoffollowingstages: narrow band (NB) colonoscopy image enhancement, binary transformation by adaptive threshold and shape recognition for obtaining pit contours. Thenthe best match for each pit contour is found by comparing each contour of the examined image with all the contours that are contained in the training sets by using Hu-moments algorithm.Finally, each pit appears outlined in a different colour according to its recognised type and each pit area is calculated.The ratio neopastic to non-neoplastic lesions is calculated for obtaining information about dangerous cases, which need immediate medical attention.The basic advantages of the proposed method are better quality of the processed image and the simple and faster algorithmfor pit pattern classification. It would help the physician by faster and more accurate recognition of colorectal cancer and to choosetherighttreatmentaccording to therespectivestage.In the paper are given also some results obtained by computer simulation of the proposed algorithm, applied on real NB images. Keywords: Narrow band images enhancement, pit pattern recognition, Kudo’s classifications system, Hu-moments algorithm. 1. Introduction Narrow band imaging (NBI) is a technique of special light observation that enhances the visualisation of the capillary network and mucosal morphology during endoscopic observation of the gastrointestinal tract.Itusesscattering and absorption properties of human tissue.The penetration depth before being scattered (partly absorbed) depends on the wavelength of the light. The shorter wavelength (e.g. blue) is reflected. Longer wavelengths (e.g. green) penetrate deeper. The image obtained through white light is a composition of slightly different tissue layers. The result is an image with focus on superficial mucous layers (blue 415 nm) and the capillary network of the deeper submucosal layer (green 540 nm). A bright but partially blurred image is the result. Figure (1) presents different wavelength light and the mucosa [1]. NBI is useful for differentiating small colorectal non-neoplastic from neoplastic polyps and is highly accurate for distinguishing low-grade dysplasia from high-grade dysplasia/invasive cancer and can thus be used to predict the histopathology of colorectal