Study In Level Set Based Segmentation Algorithms For Inhomogeneity Image

Image segmentation is an important research topic in the field of image processing,and it is also an important step in the process of target detection and recognition.Up to now,many domestic and foreign scholars have proposed a lot of image segmentation methods.However,because of the diversity and complexity of image,image segmentation is still an important and challenging research topic.Among all the methods of image segmentation,the level set model based on the theory of curve evolution is a great concern.It is based on the rigorous mathematical theory,and makes full use of the idea of the dynamic evolution of the contour curve,which can solve the problem that many excellent image segmentation methods are difficult to solve.In this paper,some classical level set models are deeply studied,and some improvements and innovations are put forward according to the defects and shortcomings.Specific research work is as follows:(1)In order to solve the problem that the edge level set model is sensitive to the initial contour,and it is difficult to segment the gray image with weak boundary.The space of the image segmentation clustering fuzzy clustering algorithm,and then using fuzzy clustering results of edge type level set evolution model is initialized,and add the distance rules to avoid the reinitialize problem using the double well potential function.The algorithm introduces the information of the image space domain,and overcomes the defect that the level set evolution depends on the initial conditions and control parameters and requires more manual intervention.Verification of breast molybdenum target mass images with very blurred boundaries.The experimental results show that the algorithm can be initialized automatically and can correctly segment the gray image with weak boundary.(2)According to the LBF model of the initial contour is sensitive and easy to fall into local optimum,this chapter proposes the introduction of a local level set segmentation global information model,the C-V model will provide global information and local information LBF model combined by image entropy,build energy functional,and gives the theoretical derivation and the numerical solution of the evolution equation of the level set.The problem that the LBF model is sensitive to the initialization and easy to fall into the local optimum is solved effectively,and the problem that the C-V model can't deal with the intensity inhomogenrity image is solved.The effectiveness of the proposed algorithm is verified by experiments on synthetic gray scale images and medical images.(3)According to the LIC model to offset image correction field no substantive constraints(bias field smoothing and slow change),leads to the bias field correction results and the image segmentation result is not very ideal,this paper proposes a local clustering optimization based on multiplication level set image segmentation model.The bias field is fitted by a set of smooth linear basis functions to guarantee the smoothness of the bias field theoretically.In this paper,the image segmentation and bias field correction are combined in an energy functional framework.The segmentation accuracy of LIC model is improved effectively and the bias field is constrained.Finally,the effectiveness of the algorithm is verified by experiments on synthetic images and medical images.