Segmentation by regions of interest of color images using self-organizing maps neural networks
No Thumbnail Available
Date
2021
Journal Title
Journal ISSN
Volume Title
Publisher
University of Oum El Bouaghi
Abstract
In many image processing applications, regions of interest are analyzed in order to extract the relevant information contained in the image. In this paper, we propose to segment regions of interest of a color image where the regions can be segmented into different number of classes. The Kohonen Self- Organizing Maps (SOM) algorithm, which is an unsupervised neural network, is used for this purpose. After, the neurons of the organized map are regrouped into clusters by using a new procedure based on the ascending hierarchical clustering which takes into account the connectedness of the neurons of the topological Kohonen map. The number of neurons groups obtained corresponds to the number of clusters in the region of interest; it can be different for each region. Experiments on different kinds of color images showed the efficiency of the proposed segmentation method.
Description
Keywords
Image Segmentation; Kohonen Self-Organizing Maps; Region of Interest