Medical Image Retrieval Based On Multi-Scales Method

With the development of the medical imaging equipment and technology, a large number of medical images like X-rays, CT images and MRI images are used in clinical medicine. By finding out the similar images with the current diagnostic image in medical image database,doctors can use these similar images to accomplish the diagnosis and treatment of the current case. The technology of the image retrieval is used in the process of the image’s searching usually. Thus, researches on medical image retrieval have important significance for clinical medicine.The main work of this thesis includes:By the research on content-based image retrieval and the theory of wavelet analysis, a novel method of medical image retrieval based on the simplified multi-wavelet transform is adopted. This method includes two processes, namely the coarse retrieval and fine retrieval.The process of coarse retrieval is accomplished based on image edge contour features. And the result of the coarse retrieval is the basis of the fine retrieval. The fine retrieval process is achieved based on the simplified multi-wavelet transform, namely on the basis of wavelet transform, the high frequency subband which contains more information is decomposed by Haar wavelet transform, then approximate information decomposed as retrieval object on the high-frequency subband is retrieved step by step. Experimental results show that the proposed method of medical image retrieval based on simplified multi-wavelet transform has a better retrieval performance.In addition, according to the multi-scale method of wavelet transform, the medical image retrieval based on multi-layer resampling template is adopted. This method uses two sampling operators to extract sampling image of each layer, and then these resampling images are used to retrieve layer by layer. Therefore, the retrieval process from coarse to fine is finished. The recall and precision on average of medical image retrieval based on multi-layer resampling template achieve a better balance, thus this retrieval method has a better retrieval performance.