When a background interpolation-judging section 107 judges that parts areas of a lower layer and an upper layer move significantly, and a visual effect can be obtained by the background interpolation, a controller 103 controls a switch 104 based on the area information of the parts area of the lower layer to perform the background interpolation. A weighted average section 106 takes an average by weighting to the upper layer and the lower layer by using the interpolated background to generate an interpolated image. By using this interpolated image, the prediction-encoding/decoding of the upper layer is performed. On the contrary, when the background interpolation-judging section 107 judges that parts areas do not move so significantly, and a visual effect cannot be obtained by the background interpolation, a background is synthesized by using the output of the interpolated image-forming section 105 to generate an interpolated image. Hence, the shape information for the background interpolation is reduced and high quality of an image can be obtained.