2.5 维帧间运动估计方法

Method of 2.5 D inter frame motion estimatio

图象点深度变化对常用的基于仿射模型的帧间运动估计方法精度影响比较大。针对这种问题，引入了特征点深度相关参数，得到一种基于２．５维模型的帧间运动估计方法。在应用于视频图象序列稳定问题时，数值实验结果证明在图象中景深变化较大的情况下，这种新方法估计的参数比基于纯仿射模型的方法准确，同时还可以得到对图象点相对深度信息（磁撞时间）的粗略估计，而计算量增加很少。

The depth variances of feature points affect the precision of inter frame motion estimation based on commonly used affine model. In this paper a new 2.5D interframe motion estimation method that considers the factor of depth changes gives a much better result than that of affine model based method with only slightly increment in computing complexity. Numerical experiment results show this improvement. Another benefit of this new method is that it gives a rough estimation of time to collide (TTC).