Automatic Feature Subset Selection for Clustering Images using Differential Evolution

Abstract

Storing and organizing huge collection of image databases is a challenge for many applications. Such huge collection of images can be organized efficiently using image content clustering. Image Clustering is mapping of images into classes according to their similarity without any prior knowledge. Clustering of images into groups can improve the efficiency of searching images in the database for various web applications. Image content characterization greatly influences the result of clustering. This paper addresses the problem of characterizing and clustering a set of images using Differential Evolution. This work proposes a new algorithm, Automatic Feature Subset Selection for Clustering Images using Differential Evolution (AFSCIDE), to characterize the images with proper selection of textural features by feature subset selection and find groups with clustering using Differential Evolution. Experiments are conducted on various benchmark datasets CUReT, UIUC.