Improvement Of The Motion Compensation Temporal Filtering

Scalable video coding(SVC) is an attractive solution for the video transmission over heterogeneous network, the storage of video on media of different capacity and the simple adaptation for a variety of display terminals. Scalable video coding only needs to encode the signal at the highest resolution to enable decoding from partial streams for different frame rate, bit rate and spacial resolution. The current most recognized framwork for scalable video coding is based on MCTF (Motion Compensation Temporal Filtering). So it is very important for scalable video coding to improve the efficiency of MCTF.In this thesis, we propose two improvements for MCTF to increase the coding efficiency, decrease the computation complexity and mitigate the block artifact in the low-passed frames. During the update step, the constraint is set that the pixels of the update block should have the uniform prediction motion vector. By this means, we can decrease the computation complexity for less update blocks on the one hand, and improve the subject quality of the low-passed frames for more accurate updating on the other hand. Deblocking the high-passed frames can introduce less block artifacts into the low-passed frames, which in turn can be taken as better reference frames for the following motion compensation. In the mean time, the deblocking of the high-passed frames can also improve the coding efficiency for the high-activitied video sequences in which block artifact is more likely to be introduced.