Image Segmentation Method Based On Operator Splitting And Graph Cut Idea

Recently,a segmentation method based on fast algorithm has attracted more and more attentions.Compared to traditional image segmentation approaches,the method has many virtues: such as the fast segmentation,the good results and the stability.First of all,the paper simply summarizes the purpose,significance and main approaches of image segmentation,and introduce and analyze Mumford-Shah model,Improved Mumford-Shah model and fast algorithm AOS,AOS-MOS and graph cut etc.Second,in the base of piecewise constant level set image segmentation method based on Improved Mumford-Shah model,in order to improve the segmentation effect and increase the algorithm speed,the paper introduce the operator splitting idea AOS,AOS-MOS into Improved Mumford-Shah model which is used in image segmentation,and then propose the segmentation methods and algorithms based on the operator splitting idea.They not only increase the image segmentation speed(the convergent speed has been increased nearly ten times ) but also improve the segmentation quality,and make the results more clear and exact.The experiments show the conclusions above.Finally,the paper introduce the graph cut idea into M-S model,and then use algorithm Ford-Fulkerson to solve the functional minimization problem by finding the minimum cut on a given proper graph.Because initial feasible flow has an important impact on the speed of the algorithm Ford-Fulkerson,this paper propose a method of giving the initial feasible flow.This method can reduce the number of steps of which are needed to adjust in the process of algorithm Ford-Fulkerson looking for augmented chain,thereby increase the speed of the image segmentation. Compared to the gradient descent method,the method of image segmentation based on graph cut idea in this paper has many virtues:such as the fewer parameter which are needed to adjust,the fast segmentation etc.The experiment shows the model is valid.