摘要 |
<p>A geometrical conversion motion compensation predicting unit (103) calculates, for every plurality of prediction modes, a motion vector and a prediction signal between a subject block existing in a subject image and a reference block existing in a reference image and having a geometrical conversion relationship with the subject block. In each prediction mode, the geometrical conversion motion compensation predicting unit (103) selects, as representative pixels corresponding to the vertexes constituting the subject block, pixels existing at those vertexes, pixels existing in the vicinity of those vertexes or interpolation pixels existing in the vicinity of those vertexes; calculates the motion vectors of those representative pixels; and calculates the motion vectors of the pixels other than those representative pixels by interpolations using the motion vectors of those representative pixels, thereby calculating the prediction signal.</p> |