Abstract: A motion compensated predicting device compensates motion by segmenting a reference image to generate an initial predicted image and estimated value correcting information. The correcting information includes area segmentation information, motion information of each area, information of each pixel not assigned with an estimated value, and information of each pixel assigned with several estimated values. An estimated value correcting device references the initial predicted image and the estimated value correcting information to decide the value of the pixel for which a unique estimated value is missing in the initial predicted image so as to finally produce a motion compensation predicted image. Through interpolation, motion vectors are decided for a portion for which estimated values cannot be uniquely determined, and hence the contour of the overall area is kept remained and distortion of area boundaries is minimized.