IJCSNS International Journal of Computer Science and Network Security, VOL.13 No.3, March 2013 64 Manuscript received March 5, 2013 Manuscript revised March 20, 2013 SOM Neural Network as a Method in Image Color Reduction Amir Maleki Anvar, Alireza Mohammadi, Abdolhamid Pilevar Department of computer science, BuAli-sina University, Hamedan, Iran Abstract Accurate and right image partitioning is one of major objectives of the various methods in image processing, specifically in medical images. Methods with full review of image areas could identify and partition available sections in an image. Due to the variety of image gray-levels, consider a method in pre-partitioning level as gray-level reduction could be so helpful. In this paper, for this step of image partitioning a neural network self-organizing map method is introduced. Competitive and single-output properties are the major reason for this self-organizing map method. Self-organizing map color reduction method is tested on human magnetic resonance imaging brain images. Human brain has five distinct areas and this method on color reduction detected all five sections and reduced gray-levels to just five levels. Keywords: self-organizing map, neural network, color reduction, segmentation, medical image. Introduction All the times researchers are trying to use more various technologies in medicine in order to recognize their limits and tissues more accurately. One of the most complex body structures is human brain. Its complexity is because of the human brain's neural network and its communication variety. Because of so much neuronal connections, its concurrency and processing speed this structure is on the way of investigators attentions. In the meantime, apart from the structure and mode of brain function, identify the precise boundaries and brain tissue has always been rather ambiguous, especially when diagnosing patient diseases is among. For this reason, many studies have been done and the various methods have been proposed. Among the methods discussed, including methods of digital image processing techniques to improve image quality and determine boundary of brain tissues. Between these methods more than separation the areas of each section, the algorithms are considered on separation of the main tissue not inside them and also less to prepare the image before applying the various methods. In this paper, introduced a neural network method especially self-organizing map as an introduction to the methods of separating areas of images and obtained very good results from the breakdown of brain tissue image. In the second part the self-organizing map neural network is briefly introduced and then in the next part the proposed method is investigated. In the fourth section, the results achieved in the implementation of this method are assessed on human brain images. 1. Self-organizing map A self-organizing map (SOM) or self-organizing feature map (SOFM) is a type of artificial neural network that is trained using unsupervised learning to produce a low- dimensional (typically two-dimensional), discredited representation of the input space of the training samples, called a map. SOM networks are different from other artificial neural networks in the sense that they use a neighborhood function to preserve the topological properties of the input space. A SOM showing U.S. Congress voting patterns visualized in Synapse. The first two boxes show clustering and distances while the remaining ones show the component planes. Red means a yes vote while blue means a no vote in the component planes (except the party component where red is Republican and blue is Democrat). Figure 1: SOM showing US congress voting results. Screenshot from Peltarion Synapse This makes SOMs useful for visualizing low- dimensional views of high-dimensional data, akin to