A method of predicting a motion vector for a current block in a current picture includes obtaining, by the moving picture decoding device, at least three motion vectors for at least three blocks other than the current block based on a direction of a reference picture, wherein the direction of the reference picture is based on a display order of the reference picture and a display order of the current picture, predicting, by the moving picture decoding device, the motion vector for the current block by using a median operation of the at least three motion vectors and decoding the current block in the current picture using the predicted motion vector for the current block.