基于图切割与C-V模型的运动目标分割

将一种基于图切割与简化Mumford-Shah模型Chan-vese模型(G-V模型)相结合的方法应用于运动目标分割中.在此方法中,利用图切割技术求解能量最优化,利用C-V模型自适应处理目标几何的拓扑变化.通过实验对此方法在图像序列中的运动目标进行了检测与分割研究.实验结果表明,图切割能量优化加速了曲线进化进程,迭代次数大大减少,同时避免了常规水平集方法中符号函数的初始化和迭代更新.对图像序列中的运动目标进行分割的仿真实验验证了该方法的有效性.

Segmentation of motion objects based on graph cuts and C-V model

The segmentation of motion objects is achieved by combining graph cuts and simplified Mumford-Shah model(C-V model).In the method,graph cuts is used to solve energy optimization and C-V model is used to process the change of geometrical topology of object adaptively.The experimental result shows that the process of curve evolution is speeded by using graph cuts algorithm and the iterations are decreased greatly.