中文摘要：

提出了一种有效的流体运动矢量计算方法．该方法利用视频流的两帧图像进行流体运动矢量的计算，目的是获得流体连续性运动的二维矢量．采用图像分析和连续性方程约束的策略进行研究，所求得的运动矢量能够反映流体运动的特点．该方法包括运动矢量的初始化、去噪、平滑和优化这4个步骤：首先，统计每个像素所在区域的直方图，利用区域直方图主要成分匹配方法完成对运动矢量的初始化；然后，对初始化结果利用多分辨率采样方法进行去噪处理；再用统计分析方法对局部模块的运动矢量进行平滑处理；最后，用流体运动的连续性方程进行约束，对运动矢量进一步优化，得到具有保持外观一致性和运动连续性的流体运动矢量．实验结果证明了该方法的有效性，对于具有多种运动的流体场景运动矢量的计算也是实用的．

英文摘要：

This paper presents an efficient method of calculating fluid motion vector. The method tries to get accurate two-dimensional fluid motion vectors through two inputting frames using image analysis methods and the continuity equation to constraint, the obtained motion vectors can reflect the characteristics of the fluid motion. The method consists four steps of initialization of motion vectors, noise reduction, smoothing and optimization. Firstly, it statistics the histogram of each region and initializes the motion vector with main components of histogram matching techniques. Then, it uses multi-resolution sampling method to remove noise from the preliminary results. In order to get accurate results while keeping appearance consistency and movement continuity, it adopts further optimization method constrained by continuity equation of fluid motion. Experimental results demonstrate effectiveness of the method. It can be applied to the motion vectors calculating in the case of multiple motions.