Image Segmentation Based On Local Prior Information And Active Contour Model

Image segmentation is a key technique and extracts the objective of interest. It is also the fundamental premise for the visual analysis. Generally speaking, image segmentation is to divide one image into some non-overlapping regions according to the intensity, color, texture and shape features.Recently, level set method has become a research hotspot in the field of image segmentation, which widens the utilization of active contours model. And it is numerically stable and capable of describing the topology change of the contour. And so it can segment the image with complex topology changes, high noise and intensity inhomogeneity. This dissertation is deeply developed around the improvement of level set segmentation method for the complicated images.The main works in this thesis can be summarized as follows:(1) An image segmentation and bias correction model based on local bias information and CV model (LBCV) is presented in this paper. A novel energy function which is minimized by the variational method is from local image information based on CV model. Because the local characteristics of the image are considered, the presented model can effectively and efficiently segment and correct the medical images with intensity inhomogeneity. At the same time, the level set evolution equation becomes an ordinary differential one, which has a good performance. Comparative experiments demonstrate that our model can more effectively obtain segmentation and bias correction results.(2) We introduce a novel model based on local statistical probability which combines segmentation with bias correction (SVMLS model). Because of the model does not give the solving results of two-phase level set and its application, we solve the two-phase level set of the model and get its evolution equation. Then, in order to segment texture images effectively, we construct an extended structure tensor. Finally, we introduce the extended structure tensor and introduce it into our energy function model to get the texture images segmentation results. The experimental results show that our method can segment medical images and texture images accurately and rapidly, and recover the bias field of medical images greatly.(3) We propose a level-set based segmentation method based on adaptive regularization for images with intensity inhomogeneity. Maximum a posteriori estimation is adopted to combine image segmentation and bias field correction into a unified framework. Within this framework, both contour prior and bias field prior can be fully used. In order to restrict bias field, we introduce an adaptive regularization. Based on this new adaptive regularization, the bias field is estimated smoother and the estimated bias field of our method introduces less structure information obtained from input image. Experimental results on both synthetic and real images show the advantages of our method in both segmentation and bias field correction accuracies as compared with the state-of-the-art approaches.