Building a Semantic Classification of Image Database from Patterns of Relevance Feedback

The representation of human perception has become one of the most active research topics in image retrieval. This paper proposes a novel search result clustering based relevance feedback mechanism for image retrieval, in which the value of image co-occurrence is used for mining the association of images and then the tolerance rough class is adapt to capturing the relationship among images in image database. Experimental results show that the performance of the retrieval is greatly improved and it is feasible to discover the knowledge in data obtained from relevance feedback by applying the rough set theory.