The following is untested, but in principle it should work (I will test it if I have more time).
You could convert the djvu file for example to jpg's like this:
#!/bin/sh
# djvu -> jpgs converter
i=1
# number of pages (392)
while [ $i -ne 392 ]
do
ddjvu -page=$i -format=pnm 1.djvu $i.pnm
pnmtojpeg $i.pnm > $i.jpg
rm -f $i.pnm
echo "page $i done"
i=`expr $i + 1`
done
(from http://caree.livejournal.com/74639.html)
Then you could use scantailor to split the pages and produce a new output (consisting of tif files).
Apply in a third step djvubind to that folder and you get your desired djvu file.