A motion prediction apparatus and method that is capable of preventing the redundancy of calculation occurring during a motion prediction. In the apparatus, a first motion estimator predicts a single-pixel motion in a top-to-top field path and in a bottom-to-bottom field path with respect to an input image by retrieving a previous image. A frame vector determining circuit determines a frame motion vector on a basis of a motion vector outputted from the first motion estimator. A scaler scales each motion vector outputted from the first motion estimator to determine a field motion vector in a different path. A second motion estimator predicts a motion in a half pixel unit by retrieving a decoded image on a basis of each single-pixel motion vector outputted from the first motion estimator, the frame vector determining circuit and the scaler. The apparatus can reduce a calculation amount required for a motion prediction in a single pixel unit to 1/4 compared with the prior art.