IMAGE RETRIEVAL BASED ON VISUAL FEATURE GRAPH

Abstract

In this paper, the content-based image retrieval is built and the retrieval efficiency is improved by proposing the visual feature graph of the image. First of all, based on low-level features and spatial relations among regions in the image, a visual feature graph of images and similarity measure between two images is proposed; Then, on the base of K-Means algorithm, several improvements to cluster the visual feature graphs are proposed based on the similarity measure. Since then, an image retrieval algorithm is done on graph clusters. To illustrate the proposed theory, we build experiments on ImageCLEF dataset and assess the effectiveness of our method and compared to the ones of work which recently published on the same dataset. Experimental results show that our proposed method is effective and it can be applied in many multimedia data systems.

Keywords: CBIR, clustering, visual feature, similarity measure.