Research Of Image Segmentation Based On Level Set Methods

Image segmentation has always been a most fundamental and important technology which is helpful for image processing turning into image analysis.It also provides the necessary preparatory work for computer vision and pattern recognition.It has been of great attention by lots of researchers.Generally speaking,image segmentation is to divide one image into some disjoint regions according to the features.The segmentation result should make these features be homogeneous in the same region and obviously distinct in different regions.All kinds of algorithms are proposed up to now.In recent years,active contour models based on level set have become the focus for many researchers.Because of the introduction of level set method,active contour model not only has perfect theory,but also shows excellent performance when dealing with image segmentation,and has been successfully applied in various fields in industries.Under the framework of active contour models based on level set,further extensions have been deeply investigate in this thesis.For the existing problems of the model,the author constructs some new models for image segmentation,analyzes the theoretical essence,and puts forward effective numerical implementation schemes.The main works in this thesis can be summarized as follows:(1)A variational level set method for image segmentation based on improved signed pressure force function combined with the statistical information of image is proposed in this paper.Firstly,a new model of active contours is constructed by using a new pressure sign function to replace the edge function.Secondly,the algorithm maintains the merits of geodesic active contour(GAC)model and Chan-Vese(C-V)model and makes the level set function stop evolution in the boundary of the target image.Finally,simulation experiments are implemented on images with poor boundaries and intensity inhomogeneity.(2)A novel level set method integrating local and global statistical information is proposed.In this method,a new signed pressure force(SPF)function is constructed by two parts.One is the global average intensity of the image,which can accelerate the evolution of the curve when the contour far away from the object boundaries.The other is the intensity average of difference image between the averaging convolution image and the original image,which can guide the evolving curve to catch the boundaries of the objects.In addition,an adaptive weighting function is utilized to adjust the ratio between the global and local terms,which can eliminate the inconvenient selection of weighting parameter.By substituting the new SPF function for the edge stopping function of the geodesic active contour model,we obtain a novel adaptive hybrid segmentation model,which is capable of segmenting the images with intensity inhomogeneity.What is more,in our method,the level set function is initialized with a binary function,which reduces the computational cost for the re-initialization step.The experimental results and comparisons with several popular models on synthetic and real images indicate that our method achieves superior performance in segmenting images with noise,low contrast and intensity inhomogeneity.(3)A novel hybrid gradient active contour model in partial differential equation for-mulation based on local intensity cluster model to correct the bias for image segmentation.The gradient descent method is a general method in image processing.Because of the traditional gradient descent method in L~2space,the convergence speed is slow for other smooth gradients by using the Sobolev gradient for the internal energy(curve length),and using L~2gradient for the external energy during the evolution of curve.The pro-posed model can effectively and efficiently segment images with intensity inhomogeneity.Experimental results obtained on synthetic and real images show the advantages of our method in terms of computational efficiency.(4)The evolution of the level set depends only on the value of level set function and the gray value of the image,which will cause too many iterations and excess evolution.The smoothness of level sets will also affect the evolving curve to find the object boundary.We propose a region-based active contour model for image segmentation.By combining the region fitting energy based on coefficient of variation with the variable exponent p-Laplace energy,the proposed method can perform well in segmenting complex images.The region fitting energy conducts the evolving curve to reach the boundaries of the objects,and the p-Laplace energy can handle the topological changes and extract the boundaries accu-rately.In order to eliminate the re-initialization step,an augmented Lagrangian method is employed to solve the optimization problem.The results of experiments on synthetic and real images demonstrate that our method can successfully segment complex object boundaries,robust to noise and less sensitive to the initial position of contours.