Probably straightforward for video/subtitle tracks at least, what seems to be needed is to create a delegate image decoder that targets the real codec and decompresses the the data in BeginBand?() before passing the call to the real decoder. Will probably need it's own "fake" fourcc like 'MvCt' or something. There are multiple compression algorithms allowed, so the real fourcc and compression method would need to be passed in an ImageDescriptionExtension probably. Note that the only common usage of compressed tracks (afaik) is in storing VobSubs with zlib.