IMAGE BASED LOGO RETRIEVAL FOR MOBILE APPLICATIONS

摘要

A new image based content retrieval system for logo and trademark identification is presented which can be incorporated in mobile services. In contrast to the traditional logo retrieval system, the main challenges include (a) the image segmentation of the target object from a cluttered background, (b) effective region descriptor invariant under the photometric and geometric transformations. Therefore, robust object segmentation is required to extract regions of interest from the cluttered background. Then a new distinctive region descriptor using the Zernike moment (ZM) phase information is proposed for information retrieval. The proposed method is compared with three major existing methods. The experimental results are presented which indicate that the proposed method has the best performance in terms of the precision and recall criterion.