| 摘要 |
A method of finding motion vectors for use in MPEG video encoding comprises generating, from a full frame, (i) a first scaled frame having a reduced number of pixels as compared to the full frame and (ii) a second scaled frame having a reduced number of pixels as compared to the first scaled frame. A first, full, best match search is of the second scaled frame is performed to identify rough motion vectors. A second best match search is performed using the rough motion vectors identified by the first search within a limited range in the X and Y directions for each macroblock of the first scaled frame to identify intermediate motion vectors. Finally, a third best match search is performed using the intermediate motion vectors identified by the second search, within a limited range in the X and Y directions for each macroblock of the full frame to identify final motion vectors.
|
