Research On Image Segmentation Algorithm Based On Level Set And Bayesian Probability Model

Image segmentation is the most basic and important step in the process of digital image processing and the basic premise of image visual analysis and pattern recognition.Image segmentation is the process of separating images into disjoint sub-regions and extracting objects of interest.Due to the close correlation between images and our life,image segmentation has always been a hot field for scholars at home and abroad.So far,thousands of segmentation methods have appeared.However,due to the influence factors such as image imaging equipment,observation angle,illumination conditions,and complex background,the segmentation of the image does not appear to be a segmentation method suitable for all images,and still faces enormous challenges.In this paper,starting from intensity inhomogeneities,weak boundary,low contrast information and bias field image segmentation,the segmentation algorithms based on level set and probabilistic statistical model are constructed.In the third chapter,we propose a novel level set active contour model based on a new composite form of signed pressure force(CSPF)function.Compared with the previous models of this type,the signed pressure force function for the proposed model consists of three terms: details information term based on the difference image;main and global structure information term based on the product image and local information term based on the original image.Integrating the different characteristics of three images into the proposed model,we can selectively use three different aspects of information to deal with the weak boundary,noise and intensity inhomogeneity images.In addition,the length term and distance regularization term information are introduced into the level set formulation.Therefore,the proposed model can ensure accurate computation and avoids expensive reinitialization of the evolving level set function.Experiments and comparisons with several well-known segmentation models show the proposed is robust to the initialization and has a satisfactory segmentation efficiency and quality.In order to solve the problems of bias field and low contrast image(especially MR image)segmentation,we proposed a probabilistic segmentation model based Local Gauss multiplicative components(LG-MC).The model can simultaneously perform bias field estimation and image segmentation.Our idea is to make use of the property that observed image can be decomposed into multiplicative components.First,the bias field representation is given by a series of smooth basic functions,the true image to be solved is represented as the function of observed image and bias field.Then,the segmentation model of Gaussian probability distribution with different means and variances is constructed by local information.Different from the existing distribution model,our model is constructed based on the local information of the true image,therefore the influence of above mentioned factors is better avoided.In addition,the fuzzifier is introduced into the energy functional to achieve soft segmentation.