Motion Estimation via Extraction of Spatiotemporal Planes and Its Application to Video Coding

运动估计是视频编码中最重要的部分之一。在这项工作中，我们试图从三维视频立方体中提取时空平面，以实现快速有效的运动估计。结果表明，与H.264/AVC相比，在相同的重构视频质量下，我们的方法减少了10%的编码时间。在此基础上，利用计算机视觉技术研究了一种有效的时空平面提取方法。

Motion estimation is known as one of the most important parts in video coding. In this work, we try to extract spatiotemporal planes from a three dimen-sional video cube for fast and efficient motion estimation. As a result, our approach reduces 10% of encoding time with the same reconstructed video quality compared with H.264/AVC. Furthermore, an efficient extraction method of spatiotemporal planes was also investigated by applying computer-vision techniques.