In other words, the B picture utilizes the following three kinds of pictures as a predicted image: 1) the I picture or the P picture of an earlier time already decoded, 2) the I picture or the P picture of a later time already decoded, and 3) an interpolated image made from 1) and 2). [0020] Thus, video decode unit 6 cannot output an image signal until a sequence header and an I picture is properl