求取真实运动矢量的一种算法

An Algorithm to Determine True Motion Vectors

针对物体基视频编码与分割应用，提出了求取特征区域真实运动矢量的一种算法。首先，利用运动检测器将图像序列的每一帧分割成静止区域与运动区域。然后只对运动区域进行运动估计。为了克服传统块匹配法的固有缺点，本文算法采用变尺寸块匹配法，大的均匀区用大方块，运动物体的边缘区用小方块；在进行块匹配搜索期间，在易误匹配的区域，为了不遗漏真实的运动矢量，先按照ＭＡＥ准则预选出多个候选运动矢量，尽可能保证真实的运动矢量包含其中，接着再根据相邻块运动矢量的相关性，从多个候选运动矢量中选出一个比较可靠的运动矢量。实验结果表明，该算法产生的运动矢量真实可靠，而且一致性好，可用于识别可靠的运动区域。

Proposes an algorithm to determine the true motion vectors of the feature regions for object based video coding and segmentation applications. Firstly, each frame in image sequence is distinguished between stationary and moving regions by using a motion detector. Then, the motion estimation is performed in moving regions. In order to eliminate the inherent shortcomings of classical block matching algorithm, a variable size block matching technique is applied in the proposed algorithm. In order to retain the true motion vectors in regions prone to mismatch, a multi candidate pre screening method is used to record all the possible motion vectors and then true motion vector is most likely selected according to the correlation of motion vectors between the neighboring blocks. The experimental results show that this algorithm may produce reliable and coherent motion vectors.