Image Retrieval using Weighted Fusion of GLCM and TSBTC Features

Innovation in imaging technology, the widespread use of smartphones and social media, along with the boost in networking and storage technology has resulted in huge image databases. Exploring and searching similar visual images has become a key topic of research. This research article presents weighted feature fusion of gray level co-occurrence matrix (GLCM) based texture features and n-ary Thepade's Sorted Block Truncation Coding (TSBTC) based color features for image retrieval. Query image feature vector is compared with dataset image feature vector. Related images having a minimum mean squared error (MSE) are retrieved. The experimental results demonstrate that the weighted fusion of GLCM and TSBTC 8-ary features with weights 0.3 and 0.7, respectively give an Average Retrieval Accuracy (ARA) of 44.74% for augmented Wang dataset and, the weighted fusion of GLCM and TSBTC 4-ary features with 0.4 and 0.6 weights, respectively gives an ARA of 74.36% for modified COIL dataset. The proposed technique performs better as compared to the existing techniques studied and proved through statistical evaluations.