A Level Set Method For Image Segmentation Of Intensity Inhomogeneities

In recent years,with the rapid development of Artificial Intelligence(AI)and computer technology,digital image processing technology has been widely used in various fields.As one of the most basic and key technologies,image segmentation technology not only provides necessary preparations for feature extraction and target recognition in the later stage of computer vision,but also serves as a key bridge connecting image processing and image analysis.Therefore,image segmentation technology has become one of the most challenging technologies in the field of computer vision processing.The main task of image segmentation is to divide the image to be segmented into several disjoint sub-regions according to different features,and extract the region of interest(ROI)from the background.With the research and development of image segmentation technology,scholars have proposed many models and algorithms in this area.Active contour segmentation algorithm is one of the most popular algorithms,and geometric active contour models which are integrated with the level set method are favored by many scholars,because It can not only realize image multi-information fusion,but also adapt to changing the topology freely.By combining the core idea of level set methods,with regard to images with intensity inhomogeneities,the paper has a thorough study and establishes a new segmentation model of images with intensity inhomogeneities based on level set method.Eventually,a large number of comparative experiments show that the proposed model can quickly and robustly segment images with intensity inhomogeneities.The main work and research contents of this paper can be summarized as the following three aspects:Firstly,the paper summarizes the basic principles of parametric active contour models and geometric active contour models.Then,the theory of curve evolution,the basic idea of level set,the initialization and numerical calculation of level set function are introduced.The paper divides active contour models into two categories:region-based models and edge-based models,and briefly describes the principles of some classical models.Finally,the advantages and disadvantages of these models in different types of image segmentation are analyzed by many numerical calculation.Second,a signed pressure force function based on Legendre polynomial is proposed.Because the traditional level set method can not accurately segment images with intensity inhomogeneities,this paper introduces Legendre polynomial into traditional signed pressure force function,and replace the inner and outer gray mean of the evolutionary closed curve with two smooth functions.Then,a local or global active contour driven by Legendre polynomials(LGLP)can be obtained.Through comparative experiments with other level set methods on various images,it can be seen from the segmentation results that LGLP model can not only solve images with intensity inhomogeneities,but also has the advantages of fast speed and strong robustness.Finally,A new signed pressure force function with parameters is proposed.In order to solve the problem that LGLP model can't segment images with multiple objects,this paper adds a parameter to control the convergence range of signed pressure force function,and combines an edge stopping function to establish LGLP~+model,which effectively improves the generalization ability of LGLP model to solve that problem.Compared with other models,LGLP~+model has better results in accuracy,iteration times and running time so that the robustness of the improved model can be qualitatively and quantitatively proved.