Improved video coding using long-term global motion compensation

We present a new approach to video coding that utilizes video analysis based on global motion features. For each encoded frame a global motion parameter set is estimated with respect to several previously transmitted frames. Using this global motion parameter set, a new motion compensation technique inspired by super-resolution mosaicing is applied. Rate-distortion optimization is used to identify macroblocks that are only affected by global motion. For such macroblocks no prediction error is transmitted. They are purely reconstructed using long-term global motion-compensated prediction, Our results indicate significant bit-rate savings compared to a state of the art H.264/AVC video codec at the same visual quality.