A MEDICAL IMAGE RETRIEVAL SYSTEM BASED ON SEMANTIC ANNOTATIONS

摘要

This paper presents the design and implementation of a semantic Content Based Image Retrieval Systems (CBIR) developed in Matlab from scratch by choosing a combination of texture, color and shape as low level features to represent the images, and by using a multilabeling classifier to associate these low level features to a semantic label. We used the Bayes Point machine classifier to classify the images. The classification results are further enhanced by using an explicit relevance feedback algorithm. The system is tested on a set of medical images combined with other types of images and the results are presented.