Multiple Viewpoints a Strategy for Multi-Step Queries Based EMD in CBIR System

Abstract

Performance of traditional content-based image retrieval systems is far from user’s expectation due to the “semantic gap” between low-level visual features and the richness of human semantics. In attempt to reduce the “semantic gap”, this paper introduces a new effective image retrieval approach-the multi-step queries strategy-that facilitates retrieval of semantically similar images from multi layers in the feature space. In the retrieval process, images are segmented three layers, raw pixel layer, blob feature layer and global image layer, and retrieval process is refined gradually. Moreover similarity between images is measured with an approximation of the earth mover’s distance (EMD), which quickly computes minimal-cost correspondences between sets of signatures. Experimental results in medical image database show that the proposed method in this paper is good enough to describe image information, and significantly improves the performance of the retrieval system.