Abstract

Keyword- H.264 Video Coding、Fast Motion Estimation、Multi-frame Motion Estimation、Long-term Memory、Sub-pixel Fast Search Algorithm

Motion estimation plays an extremely important role in the video coding. The objective of the motion estimation is to remove the temporal redundancy between video frames so that the motion compensated frames can be coded efficiently.

Multi-frame motion-compensated prediction extends the spatial displacement vector utilized in blocking-based hybrid video coding by a variable frame reference permitting the use of more frames than the previously decoded one for motion-compensated prediction. The multi-frame buffer stores frames at encoder and decoder that efficient for motion-compensated prediction. The use of multiple frames for motion compensation in most cases provides significantly improved coding gain. The multiple reference picture selection in H.264 allows the encoder store up to five previous frames for motion estimation. However, the complexity will increase five times dramatically. Numerous fast search algorithms in single-frame motion estimation are proposed in the past. As we know, a good initial point can speed up the velocity of fast search algorithm. Our proposed algorithm is focus on multi-frame condition, and provides more precisely prediction as initial point.