Satellite Image Segmentation Using Self-Organizing Maps and Fuzzy C-Means

The quality of image interpretation depends strongly on the segmentation process which is an important step in image processing. Most of the segmentation methods and approaches are not suitable for noisy environments such as satellite images of high resolution. Sometime they require a priori knowledge, and another time they do not work on all types of images. Self-Organizing Maps (SOMs) and Fuzzy C-Means (FCM) segmentation methods are widely used to process different types of simple and complex images. These two important known methods are reviewed, and summarized. In addition, a new approach is created based on SOMs and FCM. The reason for combining both methods is to create an unsupervised parameter free approach. The new approach is applied on two different types of medium and high resolution satellite images in order to examine the accuracy of the segmentation methods and the new approach. This paper and the results of experiments provide the reader with information about the improvement obtained by this approach compared to known commercial segmentation method.