Motion Vector Field Manipulation for Complexity Reduction in Scalable Video Coding

We propose a motion vector field (MVF) scaling algorithm which includes both up-sampling and down-sampling operations. Through this model, it is possible to obtain all the MVF information required for different spatial resolutions in scalable video coding (SVC) efficiently from a pre-specified spatial layer. There´s no need to go through all the computationally intensive motion estimation (ME) operations in order to gather those information. This scheme, of course, saves a huge amount of computational power since ME and rate distortion optimization (RDO) typically are the most time consuming tasks in a generic video codec. In video broadcasting application, the proposed algorithm could provide even better coding efficiency than conventional methods in the rate distortion sense.