A prediction encoding method for a motion vector according to an embodiment comprises: (a) a step for obtaining the motion vector of a target section within a frame image to be encoded; (b) a step for determining a first prediction motion vector candidate from the motion vector of one or more sections belonging to a region adjacent on the left of the target section; (c)a step for determining a second prediction motion vector candidate from the motion vector of one or more sections belonging to a region adjacent above the target region; (d) a step for selecting an optimum prediction motion vector on the basis of a comparison between the motion vector of the target section and one or more prediction motion vector candidates that include the first prediction motion vector candidate and the second prediction motion vector candidate, and outputting prediction motion vector command information for specifying the optimum prediction motion vector thus selected; and (e) a step for encoding the prediction motion vector command information.