Improved Trademark Image Retrieval System Using Relevance Feedback

Abstract

Due to increase in the competition of products and services the trademark designing has become important now a days. Therefore designing an efficient trademark recognition system is imperative. This paper proposes an automated system for trademark retrieval based on colored trademark images. Trademark retrieval system is designed by implementing techniques for color, shape and texture feature extraction. Relevance Feedback is applied to this system to improve the retrieval performance of the system. The proposed trademark retrieval approach uses Relevance Feedback and three kinds of query refinement strategies, Query Point Movement (QPM), Query Reweighting (QR), and Query Expansion (QEX). The data set consists of about 2000 color trademark images. Euclidian Distance is used for similarity computation between the query image and database images. The performance of the system is evaluated using standard evaluation parameters precision and recall. The results are compared with the conventional approach of content based image retrieval (CBIR). The relevance feedback technique has improved the retrieval performance of the system when compared with traditional approach of content based image retrieval (CBIR).