To provide a method for producing a predictive image which can improve the encoding efficiency by increasing the candidates of block images for determining predictive errors between a original image, by merely adding simple arithmetically averaging operations, and a method and device for image encoding.
Local-movement compensated block images, global-movement compensated block images, and global/local movement compensated block images are formed, by respectively using local movement compensation, global movement compensation, or global/local movement compensation. Averaged block images are formed by averaging operations on arbitrary two kinds out of the movement compensated block images stated above. One kind from among the local-movement compensated block images, global-movement compensated block images, global/local movement compensated block images, and the averaged block images is selected to determine the differences from the images in the blocks of a source image to find the differences as predictive error signals.
NAKAYA YUICHIRO