Fusion based Image Retrieval using Haralick Moments and TSBTC Features

Abstract

Due to the advancements in image acquisition devices, images have become the resourceful foundation of information representation. An enormous amount of image data is generated and shared due to the improvements in data storage and communication technologies. Searching and retrieving similar content images from these huge repositories has become a key issue. This article introduces a fusion of Haralick features and Thepade’s Sorted Block Truncation Coding (TSBTC) n-ary features for image retrieval. The proposed technique is tested on modified COIL and augmented Wang datasets. The similarity between dataset images and query images is measured using Mean Squared Error (MSE). The performance of the proposed method is tested using average retrieval accuracy (ARA). The experimental results show that the fusion of Haralick features and 8-ary TSBTC features gives an ARA of 44.20% for the augmented Wang dataset and the feature fusion of Haralick and 4-ary TSBTC gives an ARA of 74.08% for the modified COIL dataset. The proposed technique performs better as compared to the existing techniques in terms of the ARA performance metric with statistical significance.