Image Segmentation Method Based On Watershed Algorithm And Isoperimetric Theory

Image segmentation is a preprocessing stage for advanced image analysis suchas pattern recognition,its purpose is to distinguish the target from the backgroundof the image. The image is divided into some disjoint areas,the points belong to thesame area have similar properties. Because of the different characteristics ofimages, some complex image segmentation algorithm has been proposedcurrently.The article focuses on the study of the morphological watershed algorithmand graph theory algorithm, the major work is done as follows:Firstly,analyze and summarize the existing segmentation algorithm. Select theclassic watershed algorithm of the region segmentation method as a central pointand then combin some related morphological knowledge to improve this algorithm.At the same time select the graph algorithm which has been the hotspot in recentyears as the other central point,and use wavelet theory to improve it.Secondly, the morphological watershed algorithm using markers is researchedand analysised in this paper. The watershed image segmentation algorithm usingmorphological markers is proposed in this paper to resolve the problem of theover-segmentation in the watershed algorithm. This method uses the filter operatorof morphological reconstruction opening and closing to smooth the gradient image,and the marks extracted by using the H-minima technique was revised.Experiments show that the marks extracted by this method can deal with theover-segmentation problem better than tradtional methods.Finally,the cut set theory in graph theory is researched and analysised. theisoperimetric cut set theory was researched as a key point. Due to thetime-consuming problems of cut set theory, a fusion of isoperimetric theory andwavelet theory was proposed in this paper.In this method, the waveletreconstruction character and the method of using gray information to calculateequality frequency are applied to reduce the target of reducing the segmentationtime. The practicality of the algorithm is verified by experiments.