Image similarity measurement using max weighted bipartite matching

There is an urgent need for effective content-based image retrieval (CBIR) systems, and we believe the key to effective CBIR performance lies in the ability to access the image at the level of objects. Motivated by these considerations, we propose a region-based image retrieval system using max weighted bipartite matching, which can successfully solve the similarity measure of a multi-region image. The retrieval involves two stages: first, the images are segmented based on perceptual color homogeneity, for each color region, color, texture, scale, location and shape characteristics are used to represent the content of regions. Second, the max weighted bipartite matching scheme is used to measure the similarity between images. Experimental results show that a region-based approach can retrieve more relevant and more accurate images