A stream combining apparatus is provided, comprising an input unit that receives the input of first group access units and second group access units from two streams that are generated by overlap transform; a decoder that generates group frames by decoding the group access units and that generates group frames by decoding the group access units; and a combining unit that uses first group frames and second group frames as a frame of reference for the access units, that decodes the frames, that performs selective mixing to generate mixed frames, that encodes said mixed frames, that generates a prescribed number of group access units, and that joins two streams, using a prescribed number of group access units as a joint such that the access units adjacent to each other on the boundary between the two streams and a prescribed number of group access units are stitched so that the information for decoding the same common frames is distributed.