| 摘要 |
The correlation between a coded block on the screen of the current image and a plurality of candidate blocks which exist in a search area on a reference screen temporally different from the current image and which has the same area as that of the coded block is obtained to select a candidate block having the highest correlation to obtain its displacement as a primary motion vector. Then, the positional relationship between a motion vector/corresponding candidate block and the reference screen is estimated to determine whether the search area/candidate block is positioned inside of the boundary of the reference screen. When a part of the search area/candidate block is positioned outside of the boundary, the position of the primary motion vector so that the search area/candidate block is located inside of the boundary of the reference. Thereafter, a secondary motion vector is estimated using the corrected motion vector and the correction quantity to obtain a desired motion vector on the basis of the primary motion vector, the correction quantity and the secondary motion vector. Thus, the correction quantity is added during the motion vector estimation for the edge of the reference screen, to put the search area in the reference screen during the estimation of a desired motion vector, so as to improve the estimation precision.
|
