摘要
在多尺度分解的框架下,本文对不同尺度层之间的流速映射因子进行了研究.通过计算一分辨率层上估计的第(?)帧图像和真实的第二帧图像之间的最小RMSE.获得该层与前(?)低分辨率层之间的最佳流速映射因子.根据该因子映射的流速较好地反映了流速在该尺度层上的初始估计.人工和真实图像序列的实验表明,采用了最小RMSE测度,使本文给出的多尺度微分光流算法的计算精度最少提高了2.6%.
In the framework of multi-scale decomposition, the mapping(?)actor of flow velocity between different scaled levels is explored in this paper. By computing the least RMSE for a level between the estimated second image and the real second image, the optimally flow mapping factor is obtained between this level and the previo(?)s level. The flow velocity multiplied by the operator can be used as the good initial flow velocity for this level. Experimental results for synthetic and real image sequences show that our multi-scale differential algorithm proposed has the minimum accuracy inerease of 2.6 percent using measure of the least RMSE.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2004年第3期299-305,共7页
Pattern Recognition and Artificial Intelligence
基金
国家自然科学基金(No.60305003)
关键词
光流
多尺度
最小均方根误差
流速
映射因子
Optical Flow
Multi-Scale
The Least Root-Mean-Square Error
Flow Velocity
Mapping Factor
