Image Segmentation Method Based On Graph Theory

In recent years,the image segmentation method based on graph theory is a new research focus in the international community, the basic idea of which is to see the image as a weighted graph,and its every node corresponds to a pixel or region of an image.The weights to connect every two nodes show the possibility of these two nodes belonging to the same region, and the weights are related to the connections of two nodes,such as similarity,proximity,and continuity.Make the corresponding energy function according to some particular divisions,and the minimum of this energy function corresponds to one of the best sub-images.This paper first states the research status of the image segmentation and the image segmentation based on graph theory in the image technology.Because some classical image segmentation methods based on graph theory,such as Normalized cut method,have great complexity and low calculations for big-scale images,and the stability of the division greatly depends on the choice of parameters,so they can not be used in practical.Therefore,the purpose of this paper is to study an image segmentation method based on graph theory which has low computational complexity and practical values.Applying the graph theory to image thresholding segmentation can reduce computational complexity but also use the advantages of graph theory.In the reference[34],the weight formula it used contains two manual adjustment parameters,this shortage limits the generality of the method.So this paper uses an adaptive weight formula to improve the generality.A large number of experimental results show that:this new weight formula possesses a good generality and stability compared with other weight formulas.For complex image segmentation,in order to retain more image details,this paper introduces the multi thresholds image segmentation based on graph theory to solve the problem.This paper has studied the specific implementation steps and do some experiments for this method.A large number of experimental results show that:this new method possesses a good senmentation,and can retain more image details,and has practical application values.