Performance Evaluation of Image Retrieval System Based on Error Metrics

Content based image classification/retrieval system has become a widely useful component for any kind of database systems. However, the existing methods do not provide adequate information to the user query based on feature descriptors. In particular, retrieval in medical environment require special attention in feature extraction and feature selection. As, there is no benchmark retrieval algorithm proposed to a database system which contains all kind of medical images, time has come to evaluate the existing system methodology and to devise an interactive and effective retrieval system. Hence, various mathematical and human visual system (HVS) based measurement methods are applied to evaluate the relevancy between the query and retrieved images. From the obtained results and comparisons, the feature based retrieval system does not exploit in a better way to meet the requirements of medical users.