A Method of Region-based Calculating Image Similarity for RBIR System

Calculation of image similarities is one of the key technologies for RBIR (Region-Based Image Retrieval) system. In this paper, a method of region-based calculating image similarity was proposed. For this method, firstly, the similarity of two regions with the combination features (color, texture and position), histogram and region shape features, was computed, respectively. Then, the similarity of two regions was computed by the weight product method. At last, the average similarity of all regions was regarded as the similarity between two images. Three retrieval models and the implementation of image retrieval system were introduced also. This RBIR system can make person comprehend the process of image retrieval and reduce the semantic gap to some extent. It can also overcome the lack of region segmented imprecisely. The proposed method and implementation system are tested using the corel natural image database, and the results showed that the retrieval performance was greatly improved.