Image Retrieval Based on Spatial Organization of Patches

Abstract

This paper presents a novel approach for image retrieval by utilizing the spatial organization of image patches. Firstly, each image is presented as a set of patches and each patch is descripted as a bag-of-visual-words (BoW) feature. Secondly, the matrix of matching costs between all pairs of patches that come from the two images respectively is computed by utilizing both the appearance feature and spatial information. Finally, the distance between two images is obtained by minimize the total cost of matching using Hungarian algorithm. The resulting distance is applied to image retrieval. The experimental results indicate that the proposed method outperforms the traditional BoW method in image retrieval scenario.