Some ideas where nanscan could help/implement:

- The queue order of scanned documents should be kept somehow. For example,
in Open ERP even if one document has gone as an attachment to one customer
it could be interesting to know which document (page) was scanned just before
and which just after. This could be used by algorithms trying to identify
related pages. For multi page document handling we could use djvu as it seems
the right document format. It allows attaching meta information, ocr
information with characters positions is supported (perl script available to 
feed from tesseract output), it can also handle annotations.
Image compression is supposed to be very good. A first test with c44 (the 
compression application) shows results similar to JPEG.

It should be easy to attach concepts such as page, volume or collection to an image.
Templates should ease that too. So by simply setting an input as 'page' or 'volume' 
there should be some logic to handle and understand them and put them in the same 
djvu (or PDF, TIFF, whatever) document.

djvu has some interesting features regarding speed. You can even have each page (or I 
suppose a set of pages) in different files which can be queried on demand. The index
file just links to the appropiate URL for each of the pages.

- Logo/image detection/recognition

- Document segmentation

- Multiline values

- API consistency: Should we use python properties (python standard) or setProperty()
property() functions (PyQt4 standard). The latter has a couple of problems: the need 
to create both functions for each property and that if we want to embed a widget in
QtDesigner properties should be set in Python style.

- Add simple scanning dialog, with the possibility of storing the image in a file, 
server, etc.