Convert the output from "qmf -d ..." (which is a mosaic of critically decimated subbands) into a sequence, where each subband is one frame. The frames will be the size of the least decimated subband, and more subsampled frames will have pixels replicated to fill the frame.