Color Image Segmentation Using Self Organizing Map Artificial Neural Network

Segmentation of color images is one of the basic steps in most of image processing applications and quality of its result has great influence on future processes. A novel segmentation method using artificial neural network (ANN) for color image segmentation is proposed in this paper. For increasing color differences, images are represented in a modified L*u*v* color space. Self-organizing neural network with unsupervised learning is used for image segmentation based on color reduction. In color reduction, image colors are projected into a small set of prototypes using Self-Organizing Map (SOM) network. Results of this method can be considered as a near optimal segmentation with a low computation cost. Experimental results show that this system has a desirable ability for color image segmentation in computer vision applications.