摘要 |
<P>PROBLEM TO BE SOLVED: To improve coding efficiency while reducing memory bandwidth even when video frames such as macro blocks, are split into small area units to individually perform motion-compensating prediction. <P>SOLUTION: When motion-compensating prediction using a virtual sample is used in motion vector detection in the individual area, precision of the virtual sample is locally determined in relation with the size of individual motion vector detection unit area. For example, for the motion vector detection unit area of 8×4, 4×8, or 4×4 size which is smaller than 8×8 MC, a virtual sample with half pixel precision is used, and for the motion vector detection unit area with size of 8×8 MC or more, the virtual sample with quarter pixel precision is used. This allows to reduce the memory bandwidths required for the virtual sample generation for each of the 8×4, 4×8, and 4×4 MC modes and facilitates simplification of the apparatuses. <P>COPYRIGHT: (C)2009,JPO&INPIT |