Image Segmentation Based On Active Contour Model

Image Segmentation is the foundation of image processing and computer vision.In recent years, PDE-based Active Contour Model (ACM) makes a great process inthe field of image segmentation. It can effectively apply to medical image analysis,remote image processing, robot vision, video tracking and so forth, and will have atremendous theoretical and actual significance for speeding up the industrialautomatization procedure of our country, enriching and developing the existing imageprocessing methods. The geometric active contour model based on variational levelset has follow advantages: intuitionistic physical energy, stable numerical realization,can handle the topological changes automatically and use restriction informationeasily and so on. Therefore, it has applying widely in image segmentation field.This paper mainly study on image segmentation technology of active contourmodel based on variational level set, the main work is as follows:We advanced an active contour based on global fitting information according tothe piecewise constant image of intensity homogeneity through analysis the CV model.We adopt a Gaussian smoothing method to solve re-initialization problem. Themethod improves computational efficiency, keeps numerical stability, and enhancesanti-noise capacity.GIF model, LBF model and edge detection model are only using single imageinformation to image segmentation. Therefore, in order to make full use of imageinformation, we combine the three models to establish an integrated active contourmodel. The model use the global gray information, local gray information and edgeinformation, comprehensively, and also adopt the Gaussian smoothing method tosolve re-initialization problem. The results demonstrate that the combination iseffective for image segmentation of the intensity homogeneity and inhomogeneity,and show that the model is robust, highly efficiency and stable with the comparisonexperiment of CV model, GIF model and LBF model.Fuzzy clustering and variational level set both implement image segmentation byminimizing the objective functions. We introduce a variational level set model, whichintegrates fuzzy clustering algorithm through transforming the membership functionof fuzzy clustering, which combines the advantage of both methods. The experimentresults show the model can extract the target with less iterative time, handle topology changes automatically, and is not sensitive to initial contour.