Analytic Implementation for Medical Image Retrieval Based On FCM Using Feature Fusion With Relevance Feedback

A system for endoscopic image which supports image retrieval by content is described here. A new algorithm with multi-feature fusion, which referring to low-level features, is proposed based on the study of single feature including color clustering, texture information. Then, relevance feedback technique is combined with the feature fusion algorithm in order to improve retrieval results. A prototype system which supports query by example is implemented using VC# and SQL server 2000. Finally, the performance of the method is illustrated respectively using examples from an image database composed of 1361 endoscopic images, and the comparison of retrieval results shows that the approach proposed here is effective.