Abstract
This paper proposes an approach to improve the accuracy of the motion vector (MV) predictor in video coding. The proposed method makes use of the mechanism of a physical model of rigid objects to assess motion vector relationships among neighboring blocks, and takes advantage of this relationship to improve MV predictor accuracy. Consequently, more bits can be allocated for texture information to improve video quality. Our premise is that blocks associated with the same physical rigid objects have stronger MV relationships than those not associated. This association leads to certain observable behaviors. By detecting this behavior we design an algorithm to achieve a more accurate motion vector predictor. We tested our method on a multi-layer scalable video codec and the results show that the proposed method reduces the number of bits used for motion information without PSNR degradation.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call