The Medical Image Retrieval Based on the Fuzzy Feature

The global features of color, texture and shape are usually designated as the key index when content-based image retrieval (CBIR) is used in the medical field. Sometimes good result appears when these features are applied to color pathology or dermatosis images retrieval, however the topological relationships of the organs are generally ignored for most of the medical images. These disadvantages could be diminished by local features. In this paper, a new retrieval algorithm based on the fuzzy feature of medical images is proposed. Firstly, the medical images are segmented by the improved Chan-Vese(C-V) method with complex models. Secondly, in order to describe the uncertainty of the segmentation results, the features are mapped to the fuzzy feature region, and then the features of gray level, texture and shape are extracted from the segmented regions as the vision features. Finally, the comparability of the fuzzy features is calculated. The experimental results indicate that the focus distribution information of the medical images can be labelled exactly by using the proposed algorithm, which has a good robustness, high efficiency, and comparatively good recall ratio and precision ratio.