This method uses the current frame and the last two to determine whether to bob or weave a given pixel. This gives better results on both stationary and moving images than the above two methods however it uses more CPU.