Research On Medical Image Segmentation Methods Based On Mathematic Morphology

When researching on images, people are interested in some parts, which are usually with special and unique qualities. In order to distinguish and analyze those parts, we must separate them from the image firstly. To draw objective areas from the original image is defined as image segmentation. In recent years, with the development of science and information technology, image treatment has become one of the most important tools in science researches. Image segmentation, as a key technology of image treatment, has been broadly applied in military, medicine, industry, aerospace and other areas and attracted more and more researchers. Medical image segmentation is the focus of researches on image treatment recently. Segmentation based on medical images will be used to analyze medical images and help doctors diagnose illnesses and choose proper cure methods. However, medical images have the characteristics of strong noise, poor resolution, different image qualities, which makes the segmentation difficult. Thus, it is of significance to research on medical image segmentation. Focused to the characteristics of medical images, medical image segmentation methods based on mathematic morphology is proposed in this paper. This algorithm gives denoising treatment to medical images firstly, and then segment the denosing medical images, which includes two steps:medical images pre-treatment and medical images segmentation.Firstly, pre-treatment for medical images:image-forming equipment and complicated tissues and organs exert bad influence on medical images, therefore, pre-treatment should be given to medical images firstly. Wavelet threshold denoising is used as the pre-treatment for medical images in this paper. Wavelet threshold denoising (DWT) is the simplest and most usual algorithm of wavelet denoising. This paper apply this algorithm to denoise for medical image, which can have good denoising effect under the circumstance that detailed information is kept as much as possible. Medical images show the contrast of tissue density and thickness. The image element changes slightly in tissues and most medical images have big level background. Improved wavelet threshold denoising algorithm segments images firstly using Otsu and iteration algorithm into detailed area and level area, and then denoising these two parting using DWT, and at last merges two areas to obtain the final denoising image.Secondly, medical image segmentation:after denoising medical images, the referential image of mathematic morphology watershed is optimized in this paper and used in the following segmentation. This algorithm combines mathematic morphology watershed and 2-D maximum entropy threshold method solves over-segmentation of watershed algorithm while Watershed algorithm solves the computational complexity and threshold discrimination existing error of 2-D maximum entropy threshold method, therefore, medical images can be well segmented. Medical image segmentation methods based on mathematic morphology firstly reconstructs gradient image using opening-and-closing reconstruction; then obtains floating-point active-image of the reconstructed gradient image which reduce weak edges, and goes on segmentation with the floating-point active-image as the referential image; at last, after segmentation, 2-D maximum entropy threshold method will be used for fine segmentation, and better segmentation images will be obtained.In this paper, medical image segmentation based on mathematic morphology is used in the emulation experiment of Matlab. The result shows that the referential image of watershed segmentation can have good effect after optimizing.