In an apparatus that converts (compresses) motion image data, the amount of motion of a subject in a block of motion image data is detected, and spatial decimation is performed in a mode in which sampling point positions are fixed or in a mode in which sampling point positions are shifted, depending on the detected amount of motion. When predicted image quality corresponding to the moving speed of a subject detected by analyzing blocks is lower than a predetermined threshold level, spatial decimation is performed in the mode in sampling point positions are shifted such that the moving speed of the subject is virtually changed to a value at which a super resolution effect occurs, thereby achieving data conversion without causing significant degradation in image quality.