Retrieving interesting images using fuzzy image segmentation and fuzzy data mining model

We propose a novel image mining model to address three issues. First, we utilize a scalable region-based representation for an image. Users can retrieve desired images according to their visual interesting. Second, we take users´ intuition into consideration. We design a such that users can select some important portions from an image and directly search the database. Also the relevance feedback can achieve a higher performance. Third, the processes to search for interesting images can be treated as transaction records and be transformed into useful information. Through the data mining analysis, the system can discover the users´ intentions in retrieving images. The discovered rules may in turn be used to improve the retrieval efficiency. We exploit fuzzy clustering method and data mining model to design an image retrieval model. The designing methodology is illustrated and the system performance is evaluated in the paper.