Medical Image Retrieval Algorithm Based On Feature Fusion Research

With the rapid development of the research on content-based image retrieval(CBIR), the application of medical image retrieval becomes a hot research topic due to the rapid increase of out of order medical image data. This paper focuses on the analysis of medical image's content and retrieval based on low-level visual features, and develops a prototype of the medical image retrieval system (MIRS).This paper first introduces the concepts, architectures, techniques and methods of CBIR. With the introduction of HSV color space and the specialties of medical image, a feature extraction algorithm based on color statistical clustering and the corresponding similarity measure are discussed. Then in order to solve the problem of lacking the spatial knowledge of color feature, a co-occurrence matrix technology is presented to describe the texture feature of medical images.On the basis of using of color and texture features separately, a novel method for image retrieval using combined color and texture feature is presented, furthermore,-i-relevant feedback is used to gain better understanding of user's intention and high retrieval efficiency. At last, the algorithm is applied to the real medical image database, and the experimental results show that the proposed method offers significant retrieval efficiency and verify the advantage of the method based on combined featureThe result can be widely used in the clinical image management system, such as PACS, RIS, etc.