Medium having image decoding program effecting: extracting motion vector information, quantized DCT coefficients, and rounding method information; dequantizing quantized DCT coefficients to DCT coefficients; performing inverse DCT conversion on the DCT coefficients to an error image; synthesizing a prediction image of a currently decoded image via motion compensation using motion vector information, rounding method information, and a previously-decoded reference image; and adding the prediction and error images to a decoded image; wherein motion compensation is performed with half-pixel accuracy and uses bilinear interpolation to calculate intensity values of chrominance or luminance where no pixels exist in the reference image, using a positive or negative rounding method; wherein interpolation is performed using a rounding method specified by one-bit rounding method information included in a header section of coded information of the current image; wherein the rounding method specifies one of two values specifying a positive or negative rounding method.