MINING USER NAVIGATION PATTERNS FOR EFFICIENT RELEVANCE FEEDBACK FOR CBIR

Abstract

In today’s modernized world, content based image retrieval (CBIR) is considered as a bastion in image retrieval system. For making CBIR most suitable and productive technique, relevance feedback technique is used in conjunction with CBIR for producing more specific results which are obtained by taking feedback from user. However, existing relevance feedback-based CBIR methods usually request a number of iterative feedbacks for production of best search results, particularly in huge database. But this seems of no use in real world applications. In this paper, we propose a novel method, NRPF (Navigation pattern based relevance feedback method is used for enhancing effectiveness and efficiency in CBIR while copying large scale image data. In terms of efficiency, the iterations of feedback will get reduced drastically reduced substantially by using the navigation patterns discovered from the user query log. Effectiveness of our proposed search algorithm NPRF Search makes use of the discovered navigation patterns and it also produces query refinement strategies in other three kinds, Query Point Movement (QPM), Query Expansion (QEX) and Query Reweighting (QR), to converge the search space toward the user’s intention effectively. For this purpose NPRF systems are used for increasing quality of retrieved image. The experimental shows NPRF outperforms other established methods considerably in terms of precision, coverage, and number of feedbacks.