Content-Based Classification of Images Using Centroid Neural Network with Divergence Measure

The automatic classification of images is an effective way to organize a large-scale image database storing thousands of image files. In this paper, an automatic content-based image classification model using Centroid Neural Networks (CNN) with a Divergence Measure called Divergence-based Centroid Neural Network (DCNN) is proposed. The DCNN algorithm, which employs the divergence measure as its distance measure, is used for clustering of Gaussian Probability Distribution Function (GPDF) data. In comparison with other conventional algorithms, the DCNN designed for probability data has the robustness advantages of utilizing a localized image representation method in which each image is represented by a Gaussian distribution feature vector. Experiments and results show that the proposed classification model yields accuracy improvements of 5.77% and 6.97% over models employing the conventional Divergence-based k-means (Dk-means) and Divergence-based Self Organizing Map (DSOM) algorithms, respectively.