Please note the order of my list of requests:
1. extracting mapped text, in list-like format (lighter, easier to obtain by two DjvuLibre routines, but a little tricky to converto into a js object), or/and in xml format; this is really simple to get by a server routine, and difficult to obtain for a usual user:
2. cropping images (all what's needed is a tool to crop jpg already produced by mediawiki software); this is not so important.
Yes, sometimes default jpg images are not the best but IMHO a low-resolution image, cropped and self-uploaded into Commons, could be very useful as a "placeholder" to a better one coming from full resolution images of the page (usually from original TIFF or derived JP2 collections into IA).
Alex