Apparatus and method for efficient motion estimation

1. A method for processing frames of video data using inter frame prediction, the method comprising:

for each macroblock (MB) of a current frame of the video data:

during a first processing cycle, performing a first global search of MBs in one or more reference frames in a first direction to identify one or more first candidate MBs;

during a second processing cycle occurring after the first processing cycle, performing, in parallel:

a second global search of the MBs in the one or more reference frames in a second direction to identify one or more second candidate MBs;

a first local search based on the one or more first candidate MBs; and

a first determination of a sum of absolute difference (SAD) value between the MB and each of the one or more first candidate MBs;

during a third processing cycle occurring after the second processing cycle, performing, in parallel:

a second local search based on the one or more second candidate MBs; and

a second determination of a sum of absolute difference (SAD) value between the MB and each of the one or more second candidate MBs;

determining a motion vector that indicates a selected MB that is selected from the one or more first candidate MBs and the one or more second candidate MBs, wherein the selected MB is an MB that best matches the MB of the current frame of the video data and is determined based at least on the results of the first and second global searches, the first and second local searches, and the first and second SAD values; and

using the motion vector to encode the MB of the current frame using inter frame prediction.

2. The method of claim 1 , wherein the first, second, and third processing cycles are consecutive processing cycles.

3. The method of claim 1 , wherein performing the first local search based on the one or more first candidate MBs includes searching a region around each of the one or more candidate first MBs, and wherein performing the second local search based on the one or more second candidate MBs includes searching a region around each of the one or more candidate second MBs.

4. The method of claim 1 , further comprising:

performing a first sub-pixel search based on the one or more first candidate MBs; and

performing a second sub-pixel search based on the one or more second candidate MBs.

5. The method of claim 4 , wherein the selected MB indicated by the motion vector is further determined based on the results of the first sub-pixel search and the second sub-pixel search.

6. The method of claim 4 , wherein performing the first sub-pixel search based on the one or more first candidate MBs is performed during the third processing cycle in parallel with performing the second local search based on the one or more second candidate MBs and performing the second determination of the sum of absolute difference (SAD) value between the MB and each of the one or more second candidate MBs.

7. The method of claim 6 , wherein performing the second sub-pixel search based on the one or more second candidate MBs is performed during a fourth processing cycle occurring after the third processing cycle.

8. The method of claim 7 , wherein the first, second, third, and fourth processing cycles are consecutive processing cycles.

9. The method of claim 4 , wherein the first and second sub-pixel searches are each based on a half-pixel search.

10. The method of claim 4 , wherein the first and second sub-pixel searches are each based on a quarter-pixel search.

11. The method of claim 1 , wherein the one or more first candidate blocks are from a single reference frame.

12. The method of claim 1 , wherein the one or more first candidate blocks are from two or more reference frames.

13. The method of claim 1 , wherein the first direction is a forward direction and the second direction is a backward direction.