Image Segmentation Methods Based On Active Contour Model

Image segmentation is a fundamental task in image processing and computer vision.Active contour models are one of the most successful methods for image segmentation.Existing active contour models have some problems when dealing with image segmentation.For instance,the data usually contains strong intensity inhomogeneity,and interactive segmentation model usually does not consider the spatial location information,and it is sensitive to select the initial contour, and non-convexity models will lead the results to local minima. According to the above problems, this paper starts work on the CV model and region competition model, combining fuzzy connectedness, bias field estimates, global convex method and split bregman algorithm. The main work of this paper is as follows:1. An interactive method based on active contour model for image segmentation is put forward. Spatial location information is constructed through fuzzy connectedness and is adaptively fused to the improve CV model. A new convex active contour model with spatial location information is proposed by the convex relaxation method and the global optimal solution of the new model can be obtained by the split bregman algorithm. Compared with classical gradient descent methods, the proposed algorithm improves the speed of image segmentation. The contrast experiments on many color natural images show that the new method can quickly and accurately obtain the ideal segmentation results.2. A variational model combining with bias field is put forward and an effective numerical algorithm is established. It is simple and easy to implement that the model is constructed based on the logarithmic transformation to the original image, transforming the bias field into additive form. Density function estimation is built under additive logarithmic bias field. In order to enhance the noise resistance, the density function of center pixel is estimated by making full use of the information of neighborhood pixels. Level set function is adopted to deal with two phase segmentation problem when solving the model. However,level set method usually produce high computational cost and it is sensitive to initial conditions. In addition, the implementation of level set based segmentation algorithm needs to be reinitialized . Therefore, a kind of variational model with bias correction is gotten by the convexification of the above model, making it advantageous to solve the problem quickly by split bregman algorithm. To illustrate the effectiveness of the proposed method, tests are conducted on different images, comparing with the current popular algorithms.3. To segment the images with multiple components and strong intensity inhomogeneity,the two-phase model is extended to multiphase segmentation model and a multiphase global convex variational model is put forward based on bias field correction. Comparing experiments with current popular algorithms on medical brain MRI images show the superiority of the new method in segmentation precision and convergence speed.