变形运动图像多尺度级连分割方法与参数估计

Multiscale cascade segmentation of deformable image and parameters evaluation

针对变形运动图像分割中,分割模型和分割过程过于依赖分割初始化过程,并可能由此而产生较大误差的问题,提出了一种新的变形运动图像级连分割方法,并提出采用多尺度随机场的方法来对其中的背景运动参数进行估计,最终得到分割结果。根据仿真结果可以看出,本方法能够很好地对感兴趣的图像区域进行有效地分割。

The segmentation quality of this kind method strongly relies on its initialization, ff the initiation isn＇t accurate, the segmentation result will not be satisfactory. To solve th/s problem, a new deformable image cascade segmentation method was proposed. In the method, the MSILF was used to estimate motion parameters of image sequence＇s backgrounds. The simulation result shows that the segmentation method is satisfactory.