Medical Image Segmentation Based On Non-Subsampled Contourlet And Graph Theory

Image segmentation is the important part in image processing technology. With the development of the information technology, Medical image segmentation plays a more and more important role in medical applications. Medical image algorithm efficiency has important theoretical research meaning and commercial value. In medical images, there often will be unnecessary noise distribution in focal region; this will increase the difficulty of the doctor to accurately diagnosis the disease. When the medical image using a general segmentation method to segment medical image, usually need to denoise the image firstly, but the denoise progress will make the segmentation process time longer, this restrict the application of medical image segmentation technology.For the above phenomenon, this paper proposes the medical image segmentation based on non-subsampled Contourlet transform (NSCT) and graph theory. Firstly, use non-subsampled Contourlet transform the image, get the Contourlet coefficient of multiscale and directional information. Non-subsampled Contourlet transform is decomposed fast algorithm of image with the direction and with the shift invariance. Image segmentation based on graph theory is a hot topic this year, because the graph theory is a mature theory in research and application relatively, this paper mainly introduces two kinds of segmentation method based on Graph Theory:normalized cut segmentation and isoperimetric segmentation method, because the cutting method based on Graph Theory is generally a large amount of calculation, and affected by noise, so before the segmentation, we using Non-subsampled Contourlet transform extract the low-frequency information of the image, research shows that, the low frequency coefficients retain the most of the information in the image, the unnecessary noise is filtering, and it reduces the redundant information in image. So in the flowing segmentation process, we can save a lot of time, and get a good segmentation results.The proposed algorithm is verified in the specialty of medical image. The experimental results show that, the proposed algorithm obtained good results and greatly shorten the processing time, which has an important significance in practical application. The experimental results show that, the recognition algorithm has high accuracy, effectiveness and has high stability.