| Image segmentation is a classical and crucial problem in the field of computer vision and image understanding. Because of extracting only local information with disconnected boundary of the segmented region, and lack of ability to integrate prior knowledge about the segmented objects, classical no-model based image segmentation techniques cannot satisfy the requirements of comp1ex image vision app1ications.The segmentation method of geometry active contour (GAC) based on level set has an intuitionistic physical energy model and has a stable numerical realization, besides it can describe the topology change of the contour naturally and use restriction information easily, so it has applying widely in the image segmentation field.This paper mainly study image segmentation of active contour model based on level set. After reviewing the development of the active contour model, curve evolution and level set method are presented, then we proved that the level set method has a stable numerical realization and can describe the topology change of the contour naturally, in addition we introduce level set of variational method.We introduce two different GAC models in this paper: Li Chunming model which is based on gradient information and Chan-Vese model (C-V model) which is based on region information. Based on the drawbacks and advantages of these two models, we improve the C-V model by adding penalizing energy term of signed distance function, at the same time, mending the external energy term which based on region information, then the re-initialization of the level set can accomplish in the model evolution, and the degree of the segmentation result depend on the parameters decline.Finally, in this paper we propose a comprehensive optimized multi-phases C-V model, this model is based on gradient information and region information, and adds penalizing energy term of signed distance function, when used in the SAR image segmentation, the result shows that the model has a stronger robustily. |
PDF Full Text Request