Research Of Image Segmentation Algorithm Based On Mathmatic Morpholog

Image segmentation has got much more attention in field of computer vision in recent years. Image segmentation is a key step in linking up of image processing and image analysis. Because of original images are transferred to a form much more abstract and more compact by image segmentation, feature extract based on segmentation, parameter measure and presentation of target, higher image analysis and comprehend become much more likely.The descriptive language mathematical morphology used is based on the idea of set theory. It's a hot issue in the field of image processing to link up the mathematical morphology and image segmentation. Watershed is a strong segmentation tool which used the idea of mathematical morphology and has got much of people's attention.This paper discusses some basic concepts and theories of image segmentation and mathematical morphology, analyzes respectively the segmentation skill based on edge detection and segmentation based on region. The major key point is to research the open and closure operation of mathematical morphology, and the extend contents of expansion and corrosion operation and their properties. This paper introduces from binary morphology to the subject of grayscale image. Watershed is a generally used image segmentation tool recent years using the idea of mathematical morphology. This paper discusses researches the merit and demerit of watershed method when used in image segmentation, discusses the appearance of over-segmentation, and analyzes the reason of the over-segmentation.Specific to over-segmentation of the traditional method, this paper designs an multi-shape multi-scale structure compound filter using morphological structure elements to filter the noisy in images and retain unbroken detail information of images. Then use the traditional watershed method on the gradient images, as a preprocess step before watershed transform.The over-segmentation could be effectively solved using this improved watershed segmentation algorithm is approved through sufficient experiments, and results of which make it clear that it has distinct improvement in elapse time, contour extraction, noisy jamming and over-segmentation when comparing with the traditional watershed algorithm.