Information Technology Reference
In-Depth Information
immediate lower level. Sum of Absolute Difference (SAD) is used for cost
function. The vector which gives the minimum cost is recorded as best MV and
the cost is recorded as best cost. The search pattern used at each level is square
search with search range of ±5, ±4, ±3, ±2 and ±1 at levels 4,3,2,1 and 0
respectively in CIF. Thus there are at most 121, 81, 49 25 and 9 search points for
each candidate list at levels 4,3,2,1 and 0 respectively.
Fig. 13 Dirac Hierarchical Motion Estimation levels for CIF video format, where w is
search range
5.3.2.2 Refine Pixel Accurate Motion Vector to Sub-pixel Accuracy
Pixel accurate motion vectors are refined to sub-pixel accuracy. Dirac provides
motion vector accuracy upto ⅛ pixel, but ¼ pixel accuracy is default. Sub-pixel
refinement process uses the pixel accurate motion vector as the initial guide for
finding ½ -pel accuracy motion vectors. Similarly ½-pel accuracy motion vectors
are used as a guide for ¼ -pel accuracy and ¼ -pel accuracy motion vectors are
used as a guide for ⅛ -pel accuracy motion vectors, as illustrated in Figure 12. In
order to achieve sub-pixel accuracy, the reference frame is up converted by 2,
multiply the pixel accuracy motion vectors by 2 and search around the square
window of ±1 to get motion vectors of ½ accuracy and so on.
Once pixel-accurate motion vectors have been determined, each block will have
an associated motion vector (V0, W0). 1/2-pel accurate vectors are found by
finding the best match out of (V0,W0) and its 8 neighbours: (V0+4,W0+4),
(V0,W0+4), (V0-4,W0+4), (V0+4,W0), (V0-4,W0), (V0+4,W0-4), (V0,W0-4),
(V0-4,W0-4). This in turn produces a new best vector (V1, W1), which provides a
guide for 1/4-pel refinement, and so on until the desired accuracy. The process is
illustrated in the figure below.
Search WWH ::




Custom Search