A novel approach to self order feature reweighting in CBIR to reduce semantic gap using Relevance Feedback

Abstract

Content Based Image Retrieval (CBIR) is a prominent research area in effective retrieval and management process for large image databases. Which was a bottleneck in reducing semantic gap issue to solve, many approaches have been proposed. Among them, Relevance Feedback (RF) is a technique absorbed into CBIR systems to improve retrieval accuracy using user given feedback One of the traditional methods to enact relevance feedback is Feature Reweighting (FRW), it is useful technique to enhance retrieval performance based on the acquired feedback from user. The assumption for previous FRW approaches are that the length of feature vectors for images are fixed and use only the information from the set of images send back in the early query result for feature reweighting. In this article, we examined systematically the proposed system with various weight update strategies and compared output retrieval results and proposed a new self order feature reweighting approach in CBIR to reduce semantic gap using relevance feedback which we experimented with COREL database with 25 different categories and each category containing 100 number of relevant images. The experimental results demonstrated the advantage of our method in terms of precision and recall. The results show the success of the proposed approach and it is shown that our perspective outperforms previous work.