| 摘要 |
PURPOSE:To perform prediction in more detail than ever and to improve prediction performance by applying motion compensation to the victinity of the boundary between an object in which a motion prediction error tends to be generated and a background to each small block. CONSTITUTION:Two dynamic images continuing timewisely are divided into plural hexagon blocks 160, and 24 triangle blocks are generated in them. Thence, the presence/ absence of an intersection between the boundary 170 of a photographic object and the background and each side of the hexagon block is checked, and an apex 172 provided at the side possessing the intersection 174 is moved to the intersection 174. Thence, the intersection between six straight lines connecting an object apex 180 to six apexes 182 in the periphery of the apex 180 is found, and distance between an objected intersection 184 and the object apex 180 is calculated, and the object apex 180 is moved to a point at the minimum distance. Such processing is applied to seven apexes 164, 168 in the block, and the triangle block in a hexagon can be deformed. In this way, the motion compensation on the hexagon block including the boundary between the object in which the prediction error tends to be generated and the background by the blocks smaller than every, which improves the prediction performance. |
