Research Of Image Segmentation Algorithm Based On Graph Cuts Theory And Laplace Level Set Algorithm

As a key step in image analysis and image understanding,image segmentation has been attached much more importance by people,the high-precision image segmentation algorithm can provide reliable and scientific data for the next image analysis and image understanding.However,because of the difference of image information and the needs of people,there has been no unified segmentation method can give a general solution for all of the questions,researchers have developed many different segmentation algorithm to adapt to solve specific problems,which caused the diversity of image segmentation algorithm.The same time,because of the complexity of the internal structure of some images,the segmentation algorithms become difficult.In recent years,graph cut algorithm based on combinatorial optimization from graph theory and level set algorithm based on curve evolution theory are widely used in image segmentation.These two algorithms are both based on partial differential equations,we can build mathematical model which correspond with our questions and establish suitable energy function and solve the partial differential equation,so as to we can obtain the optimal solution of segmentation results through making the energy function minimized.Graph cut algorithm is base on graph cut theory,it can get the global optimal solution by optimizing the energy function defined discrete variables.Mean while,level set algorithm is based on the curve evolution model,in order to get the optimal solution,this method drive the initial closed curve to gradually approach the outline of the divided region through minimizing the energy function.This paper mainly focused on the two algorithms above,the specific content including as follows:(1)On one hand,this paper has researched the theoretical background of graph cut algorithm and elaborated the definition about graph cut theory:network flow,cut,maximum flow,minimum cut.After researching the concept of the energy function,firstly,we build the mathematical model according to the region information and edge information of image,secondly,we use labels to mark the target and background region of image,thirdly,we establish the energy function of graph cut to transform the object segmentation problem into energy function minimization problem.Finally,we can get the global optimum or the local minimum which approaches the global optimum of initial image.In addition,this paper also proposes an improved algorithm based on mean shift algorithm and graph cut algorithm,the first step is image pre-processing,then we take every sub-region as a network node to construct the network graph,finally we prove the effectiveness of the improved algorithm by experiments.(2)On the other hand,this paper deeply study the theoretical basis of level set algorithm,A new image segmentation algorithm based on the Lapalacian Level Set is proposed in this paper,this algorithm combines regional information into speed function to drive the evolution of level set surface.The algorithm utilizes not only the information of image edges and gradient information,but also image region information.The algorithm takes advantage of regional global optimization features meanwhile maintaining the local features of edges.The new proposed algorithm implements effective segmentation of images.The improved algorithm is implemented under the ITK(The Insight Toolkit)platform.The experiment results indicate that the improved algorithm has good performance in maintaining the continuity of the edges,so that the segmentation result is relatively complete.This algorithm can provide reliable scientific data for image analysis.Finally,we summarize the algorithms of this paper and analyse the advantages and disadvantages of the algorithms,then we make a future thought about the next job.