Image Distance Measure Based on Adaptive Patch Matching

This paper presents a novel approach to measuring distance between images and exploit it for image retrieval. Firstly, each image is presented as a set of patches, integrating both the local appearance feature and spatial information of patches. Secondly, the set of matching costs between all pairs of patches of two images is computed. Then the distance between two images is obtained by minimize the total cost of matching using Hungarian algorithm. Finally, we test our image distance measure method in colorful image retrieval. The experimental results indicate that the proposed method outperforms other existing image distance measure methods.