Research On Method Of Brain Tumor Segmentation Based On Morphological

Brain MRI and CT scan can provide reliable diagnostic basis for clinical doctors, owingto its quick and easy maneuverability, effective and accurate diagnosis. The main purpose ofthis issue is to research on brain tumor segmentation algorithm based on the characteristics ofbrain CT images. The main goal is to realize the automatic segmentation method of braintumors by computer without artificial intervention. on account of the complicated andrandom-shaped structure of brain tissue, as well as the differences among individuals, whichmakes the segmentation algorithm large amount of calculation and prone to error, alsoreduces the precision and accuracy of segmentation results. This paper focuses on watershedsegmentation algorithm and morphological operations, furthermore, it also studied that themathematical morphology multiscale correction and viscous morphological operator caneffectively restrain the over-segmentation. The main work are embodied in the followingaspects:1. Different from traditional morphological opening and closing operation, we proposeda multi-scale correction method, which determine the size of the structural elements based onthe mean square error between different pixel gradient value and the gradient value within itscertain neighborhood, modify each pixel with morphological opening(closing) operator. Bythis way, most local minimums caused by irregular details and noise were removed, andregion contours positions were not changed or slightly changed.2. In consideration of parameter morphological watershed segmentation, we bring theviscous fluid into watershed process. Thus, most local minimums caused by irregular detailsand noise were flooded with more viscous fluid, whereas the less viscous to the larger targetareas. In this way, it can produce closed object contours, on the other hand, it can also avoidthe over-segmentation phenomenon.The experimental results indicate that the proposed algorithm can effectively avoid theover-segmentation in of watershed and also precisely locate the region contours positionsunder the condition of pockety gray and fuzzy boundaries.