You probably have to roll your own version if you want it. http://answers.yahoo.com/question/index?qid=20090421184939AA...

I haven't looked too far into this domain but I would imagine that it is not too difficult to do on a smallish scale.

On the actual form itself have some kind of scale and position marker in the top left and bottom right corners to give you orientation and scale of the scanned image. Then do some image analysis at predefined positions on the page. (of course you need to take into consideration the scale of the sheet and the orientation to make sure your offset vector is going in the right direction and distance per question.)

edit: this may be what you are looking for:

http://www.cs.uwaterloo.ca/~a3seth/udai/OMRProj/

Thanks. This link looks useful.

its not quite what you're looking for, i think, but i've heard good things about http://www.pdftoword.com/

Looking interesting. But as you guessed, I am not looking for OCR (optical character recognition).

I am looking for capturing which circle was filled in: A or B or C.

well, that service does more than that, including trying to capture and recreate images.

yes, ... except I want to move beyond the "image" of circle A shaded in, and circle B not shaded in.

I want a CSV text file listing the captured data: 1,A 2,B 3,B 4,A etc.

while, again, i did mention that this isn't the out-of-the-box solution you're looking for, i think it could provide you with a good place to take off with something custom written.

Doing a little more online research, it seems like I am looking for a free/open-source solution for OMR (optical mark recognition) using a generic image scanner or PDF (instead of a specialized device that just scans OMR forms).

I wonder if there are computer/machine vision hackers who might know where to find (or put together) a good solution.