Query modification from user relevance feedback by multiple alignment for image retrieval

This paper presents a query modification from user relevance feedback by multiple alignment for image retrieval. In this paper, we propose a novel idea of query modification by aligning multiple positive and negative feedbacks. A object-based image retrieval system is also implemented to verify the effectiveness of the proposed method. Although the initial estimate of user perception is learned from the user feedback, the proposed system automatically modifies the query in question such that the new query agrees (disagrees) with the optimal set of common characteristics of the positive (negative) feedbacks by dynamic programming. The goal of the proposed multiple alignment technique is to move a query from its original position to a meaningful position in the feature space according to user relevance feedback. Experimental results on a large collection of images have shown the effectiveness and robustness of t he proposed algorithm.