Research On Improved Image Segmentation Methods Based On Level Set

Image segmentation is a crucial preprocessing for image recognition and computer vision,which refers to the technology and process of dividing an image into several specific regions with unique characters and extracting the targets of interest.The segmented sub-regions show the consistency in some cases,while the adjacent regions show the obvious differences.According to the applied requirements,the image segmentation method is different in the different types of images.Until now,a general segmentation method has not been found.At present,most of the image segmentation algorithms divide the image into regions with the certain characteristics according to the grayscale,color,and texture of the images.Then,the target regions of the interest have been extracted.However,the inhomogeneity,noise,weak boundary and low contrast of the real-world images increase the difficulty of image segmentation.The level set-based image segmentation methods are based on some mathematical theories that have high efficiency numerical analysis and computational ability,and they can flexibly deal with the topology change of target objects.Recently,the image segmentation methods based on the level set are a relatively active research direction.Based on the traditional active contour model,this paper improves the image segmentation methods based on the level set.It follows that two image segmentation methods based on the level set are investigated.The effectiveness and feasibility of the proposed methods are verified by the simulation and quantitative evaluation experiments.The main work of this degree thesis is summarized as follows:(1)When the level set algorithm is used to segment an image,the level set function must be initialized periodically to ensure that it remains a signed distance function.In order to solve the problems of computational complexity and time consumption caused by periodic initialization,an improved regularized level set method-based image segmentation approach is presented.First,a new potential function is defined and introduced to reconstruct a new distance regularization term to solve this issue of periodically initializing the level set function.Second,the new distance regularization term is combined with the internal and external energy terms based on the image information to form a new energy functional.Finally,the partial differential equation corresponding to the energy functional is obtained by using the calculus of variations and the steepest descent approach,so as to realize the effective image segmentation.Numerical experimental results demonstrate that the proposed method is effective and robust to segment intensity-inhomogeneous and noise images.(2)When the existing algorithms are dealing with the intensity-inhomogeneous images,they tend to pay too much attention to local image information,which leads to a poor general performance of the algorithm.To solve this problem,an image segmentation method using a novel active contour model that is based on an improved signed pressure force function and a local image fitting model is proposed.First,a weight function of the global grayscale means of the inside and outside of a contour curve is presented by combining the internal gray mean value with the external gray mean value,based on which a new signed pressure force function is defined.The function can segment blurred images and weak gradient images.Then,the local image fitting model is introduced by using local image information to segment intensity-inhomogeneous images.Subsequently,a weight function is established based on the local and global image information,and then the function is used to dynamically adjust the weight relationship between the local information item and the global information item,thereby achieving the purpose of accurately segmenting the image.The experimental results show that our model involves simple computation,exhibits fast convergence,and can effectively segment multi-objective images and intensity-inhomogeneous images.Furthermore,the proposed method is highly robust to the initial contour and noise.(3)In order to further verify the segmentation performance based on the improved regularization level set algorithm,the algorithm is applied to deal with the clinical medical image segmentation problem.The simulation experiments are carried out on 4 medical images.The experimental results verify the effectiveness of the improved algorithm,which indicates that the algorithm has good performance in dealing with medical image segmentation.