Abstract
The objective of this study is to develop and examine the performance of an image classification system using Fuzzy C-Means (FCM) on a large set of images represented by MPEG-7 low-level descriptors. This experimental data set consists of five different categories of images. In a series of experiments we considered 5 different categories of the MPEG-7 descriptors related to colors and textures of images. Prior to any clustering the original space was reduced using the standard Principal Component Analysis (PCA). A series of carefully organized experiments has led to a number of interesting findings as to the suitability of fuzzy sets in the framework in image organization and description, insights into the structure of various categories and their interrelationship.