Using evolutionary graph for image segmentation N. AMMOUR 1 , A.GUESSOUM 2 , D. BERKANI 3 . 1, 2 Department of Electronic, University of Blida, Road of Soumaa, PB 270, Blida, ALGERIA nassimammour@yahoo.fr guessouma@hotmail.com 3 Department of Electrical Engineering, National Polytechnic School of El Harrach, Road of Hassan Badi, PB 182, Algiers, ALGERIA dberkani@hotmail.com Abstract: - This paper presents a new method of no supervised image segmentation. It rests on an original strategy which consists in making progress an evolutionary graph which composes the segmented image. The evolution injunction is established statistically after the crossing of a region. The matrix of space composition of the areas in each class is then given. A map of space delimitation of the regions is established by a new way of contours localization and refinement. At last, the segmented image is erected by the combination of the chart of contours and the matrix of the regions. Key-Words: - Evolutionary graph, transition, node, class, contour, region. 1 Introduction The segmentation of image is a pre-treatment, which improves the state of the information contained in the image before the desired treatment and application. The objective is to separate, in the most faithfully possible way, the objects and the bottom which make the image [1] [2]. Image segmentation has applications in many practice fields, it has outlets in the forms recognition, the objects detection, the analysis of medical image, robotics [3], or in the field of the images by satellites and well of others still. Several developed techniques of segmentation are based on a preset number of classes in the initial stage of the algorithm, which ensure the classification of each pixel of the image in its most probable class. Segmentation by region based approaches, the segmentation by contours detection, segmentation by thresholds, and that based on the method of the k-means. Classification by the theory of the obviousness, also called Dempster-Shafer theory or theory of the belief functions, it makes it possible to process on the one hand dubious data and on the other hand to combine information coming from several sources, before the use of the decision rules for the assignement class selection [4] [5] [6]. Classification by the hidden chains of Markov [7] [8]. Bayesian Classification which is based on the determination of the conditional probabilities to estimate the membership of an individual to each class [9]. Fuzzy classification [10]. This paper presents a new method of image segmentation based on the evolution of a graph, which traces the regroupings of the occupants of an image. In a first place, the methodology, which is articulated around two principal phases, is represented in the second section. The stage of creation of the classes graph is first of all carried out; it is illustrated in the third section. The wiliness in this phase is to push back the decision-making at the end of the crossing of a region. At the end of this stage, the border of each area is vague. In the third section the stage of a contours map creation is represented. The role of this phase is well to detect the borders of each region. Proceedings of the 11th WSEAS International Conference on SYSTEMS, Agios Nikolaos, Crete Island, Greece, July 23-25, 2007 131