Comparative Study of Color and Texture Features for Image Retrieval on Natural Datasets

Abstract

The size of image datasets is increasing with a rapid pace having many application areas. To search and retrieve an image with its best matches, one needs to apply different image retrieval algorithms. The popular algorithm for retrieving images based on low level visual features is known as content based visual image retrieval system. These algorithms are distinctly dependent on their capability to extract low level features and their similarity matching mechanism for getting similar images of high accuracy and precision. This paper compares some latest methods of feature extraction having texture and color features based on their retrieval performance. The similarity into images is calculated using several distance based dissimilarity metrics. A GUI based software application (written in MATLAB) is built for comparison and evaluation of retrieval performance.