Comparison Study Of Local Pattern Algorithms In Content-Based Image Retrieval Systems

Abstract

The increasing need to retrieve images from huge databases has made image retrieval system an imperative and necessary field of research. Researchers have proposed a lot of image retrieval algorithms by extracting important and distinctive features from the visual content of the image for the importance of the extracted features in improving the accuracy of CBIR systems. In this paper, a study has been made to compare the effectiveness of six famous local pattern algorithms: LBP, LTP, LTrP, MMCM, COALTP and LMP and testing these algorithms using two different types of databases: color image databases and texture database, using four distance measures (L1, Euclidean, Cityblock and Cosine) to retrieve the images with the shortest distance. The performance of the studied algorithms was evaluated using three measures: Average Retrieval Precision (ARP), Average recall and Average Retrieval Rate (ARR). This study revealed the superiority of COALTP over other tested algorithms. In addition, the results showed that local pattern algorithms were more efficient in retrieving images from texture databases as compared to color databases.