| 摘要 |
A block matching motion estimation algorithm uses reduced precision data when comparing blocks in the pictures to be compressed with blocks in a reference picture. This yields a plurality of best match candidates (66). The average (70) of the best match candidates (66), excluding those candidates (68) which fall outside a predetermined cluster boundary (74), is determined and then a second search is performed focused on this average (70). In this way, the amount of hardware for the block matching algorithm is decreased and the speed of its operation is enhanced. |
