A Relevance Feedback Image Retrieval Scheme using Combination of Color and Shape Features*

In content-based image retrieving, studying focuses on how to retrieve image to satisfy the user quickly and accurately. Color and shape features are usually presented by color histogram and edge histogram in HSV color space. But in this paper, the combination of color and shape features has been used in the feedback. In matching of similarity computing, the Gaussian model is used to normalize the different sub-characters distance. This distance is preparing for retrieving integer by combining different sub-characters. And with relevance feedback mechanism, the satisfy retrieval results can be gained by adjusting weights. Experiments showed that this method is effective in image retrieval. Keywords: Color histogram, Edge histogram, Feedback, Feature combination * Supported by the National Natural Science Foundation of China(Grant:60234030); the Foundation of CoSTIND(A1420060159)