Multiscale Cascade Segmentation of Deformable Image and Parameters Evaluation

Methods based on deformable model have been widely used in deformable image segmentation. The segmentation quality of this kind of methods strongly relies on its initialization. If the initiation isn´t accurate, the segmentation result will not be satisfactory. To solve this problem, we propose a new deformable image cascade segmentation method. In the method, the MSRF and SMAP will be used to estimate motion parameters of the background of image sequence. From the result of simulation, we can conclude that the segmentation method is satisfactory.