| 摘要 |
<p>PURPOSE: To minimize prediction errors within the range of a bit budget assigned to a motion vector by selecting the motion vector based on a prediction error, produced by a local area of a coded image. CONSTITUTION: A motion vector compression device 16 selects a motion vector among motion vectors generated by a motion vector generator 13 and selected by a motion vector-selector/coder 14. The selection is started by arranging motion vectors with respect to a large block in the order of magnitude of prediction errors, and ech motion vector with respect to the large block is replaced with sets of motion vectors of a small block corresponding to the large block, while searching the sequence list in a descending order. The replacement is continued, as long as a bit budget is not exceeded. Thus, an area using a small motion block or a large motion block is selected to minimize the prediction error, within the bit budget assigned to the motion vector.</p> |
