To achieve an encoding system including a highly efficient prediction performed in response to the content of a scene, a significance, and a motion characteristic of a moving picture and the like, memories a, b, c, motion compensator 5 responsive to an arbitrary transform parameter representing the motion of a prediction picture segment for generating a predicted picture by using arbitrary data stored in the memories a, b, c based upon the transform parameter, and memory update unit 15 for allowing the content of one or more of the memories to be updated at an arbitrary period of time, are provided.