Frankly, in the long run, to avoid conversions and giant image folders (my DR1000 has in tens of thousands images by now

), it seems to me the best solution would be to port an image viewer...
Or even better a comic book reader like
Comix, a GPL'd Python+GTK application that offers things like reading straight from compressed image folders and 2-page views, which are things requested already by the community. It would be a fantastic complement to FBReader's functionality.