A heuristic search method of adaptive interpolation filters in motion compensated predictive video coding

The paper studies the fractional pixel based motion compensation for video coding where fractional pixels are generated by filtering with adaptive interpolation filters. The design of adaptive filters is formulated as an optimization problem where the prediction error is minimized for each individual frame. A heuristic search method is introduced to obtain an optimized filter with high efficiency and accuracy. Performance of the method is analyzed and examined through simulations. The new filter design method saves up to 75% computation time while retaining coding efficiency equivalent to state-of-the-art methods reported in the literature.