Modified Feature Descriptor for CBIR using Weighted Adjacent Structure

Abstract

Content-based image retrieval (CBIR) has grown the most influential research directions in the area of digital data management.  Several researchers are trying their hand in this research application because of its appearance in many territories.   In this research, a novel feature descriptor CMS is presented.    For the texture extraction, shape, and color, Micro Structure Descriptor (MSD), Saliency Structure Histogram (SSH), color Difference Histogram (CDH) are combined. And in the end, Weighted Adjacent Structure (WAS) is practiced for the verification of the accuracy and the efficiency.  Many experiments have been performed on different query images from the Corel dataset.  For the execution, parameters, such as Recall, Precision, Time cost and F-measure is computed.