Variational Level Set Models For Image Segmentation

With the rapid development of the information and computational science, thereare a large number of digital images in applications. How to take an effective and quickway to extract useful information, e.g., image segmentation, becomes a very importantissue. Partial differential equation (PDE)-based method (geometric active contourmodel) is a kind of effective method of image segmentation. The basic idea is to deforma curve, surface or image under a partial differential equation (PDE) with initial andboundary conditions, and obtain the desired segmentation results as the solution of theequation. If the evolution PDE of level set function can be obtained from minimizationof an energy functioned defined on the level set functions, then this method is known asvariational level set method.Our studies are concentrated on variational level set method: Made the followingresults:1. Study on edge stopping functions of edge-based active contour models. Weproposed an adaptive varying stopping function.The edge stopping function is very important in edge-based active contour models.In the edge active contour model, stop functions directly influence the ability and thedivision of the model results. It is typically the composition of a strictly monotonicallydecreasing positive function and the gradient magnitude of Gaussian smoothed image.The active contour models based on this type of edge stopping functions have thedrawbacks of high sensitiveness to noise and inaccurately locating the edge of theimages with intensity inhomogeneity.In this paper, a new adaptive varying stopping function is proposed to overcomethe two drawbacks above.2. Study on the adaptive distance preserving level set evolution model. Weproposed a selectivity adaptive distance preserving level set evolution model.Adaptive distance preserving level set evolution model is derived from level setevolution without re-initialization model, which introduces a variable weight coefficientand so eliminates the need of initial contours. However, this model has the drawbacksof high sensitiveness to noise and locating inaccurately the object edge of image due tointensity inhomogeneity. Following this model, this paper introduces a new variableweight coefficient and a new edge stop function based on this variable weight coefficient. Experimental results show that the proposed model can really overcome theabove-mentioned drawbacks.3. Study on medical angiography images segmentation. We proposed a secondorder edge detection level set model.The extraction of medical Angiography image is a key step. Blood vessels haveintensity inhomogeneity, fuzzy, noise and low contrast characteristics, these featuresmake traditional active contour models difficult to extract blood vessels in medicalAngiography images. We propose an a second order edge detection level set model toaddress this problem, in which a novel variational formulation (external energy) for thelevel set function is presented. Experimental results shows that the proposed model cansegment the medical angiography image efficiently.