| 摘要 |
The invention relates to a process for the differential coding of a motion vector (MV0) associated with a macro-block (M0). This process comprises the steps: determining the components (MV0x, MV0y) of the said motion vector (MV0), determining a prediction vector, each component (Px, Py) of which is equal to the median value of the corresponding components of at least three candidate motion vectors (MV1, MV2, MV3), a candidate motion vector being a vector associated with a previously coded macro-block , subtracting the components thus obtained from the components of the vector to be coded. The invention also relates to a process for decoding motion vectors coded in the manner described above, as well as to the implementation devices. The invention applies to the field of image compression. |
