Adaptive motion model selection using a cubic spline based estimation framework

A block based video coder that supports multiple motion models is proposed. Apart from the typical translational motion model, we employ parametric models to more accurately represent complex motions that occur in video sequences. A novel method for estimating the warping parameters in a rate-constrained way is presented. A cubic spline framework is utilized to obtain fractional-accuracy samples for motion compensation. Efficient motion vector prediction schemes are developed to maintain the continuity of the predictor in spite of different motion models. Bit rate savings up to 11.7% in IPPP and 9.3% in Hierarchical B pictures are shown compared to an improved H.264/AVC reference.