基于多约束融合的二维速度场估计

2D MOTION VECTOR FIELDS ESTIMATION BASED ON MULTICONSTRAINTS FUSION

从速度场的基本约束方程出发,提出了一种多约束协同工作的二维速度场估计方法。并利用马尔可夫随机场理论进行约束的融合,从而将速度场估计转化为基于贝叶斯方法的统计决策问题。最后引入"平均场"理论进行解的优化。实验结果表明,多约束的介入能够增强算法的鲁棒性,从而得到更为准确的解。

This correspondence put forward a method for motion vector fields estimation based on multicon-straints. MRP theory provides the framework for constraints fusion and changes the motion estimation into statistical problem. Finally, ' mean field ' theory is adopted in the optimization procedure. Our experiments show that modeling with prior knowledge and optimizing with ' mean field ' theory enhance the robustness of the method and improve the accuracy of the solution.