To improve predicative performance of an entire frame with respect to a method for coding a dynamic image.
A global motion compensated predictive picture 7 is supplied to a weight average calculating part 12 along with a local motion compensated predictive picture 11, and a mean predicted picture 13 being a weighted mean of them is generated. The global motion compensated predictive picture 7, the local motion compensated predictive picture 11 and predicted picture 13 by the weighted mean of them are supplied to a predictive mode selecting part 14, and a predicted picture in which a predictive error with the input picture 1 is to be minimum is selected from among the global motion compensated predictive picture 8, the local motion compensation predicated picture 11 and the predicted picture for compensation of the mean predicted picture 13 by the weighted mean, for each small area.
KAMIKURA KAZUTO
WATANABE YUTAKA