预测通常与边界数据有关，例如用于运动补偿的运动矢量。
Prediction usually involves side data, such as a signal telling the decoder a motion vector to use for said motion compensation.
而通过采用相关计算进一步细分灰度投影数据，可以提高图像运动特征位置信息的精度，获得亚像元级的图像运动矢量。
Using correlation calculation to further subdivide gray project data, precision of image motion characteristic position information was improved, and sub pixel image motion vector could be gotten.
经过优化，对于插值过程本身，以及到后来计算运动矢量时的数据缺失引起的CPU数据挂起，都降低到了比较理想的水平。
After optimizing, the process of interpolation and the CPU data stall aroused by data missing of motion estimation are both decreased to an acceptable level.
应用推荐