(Archived) Required / recommended image resolution?

[doomedtx 0]

I've searched and not found an answer to this question--is there any information on what resolution is needed for proper text recognition in an image?

My wife has 117 JPGs of various magazine clippings, etc from an earlier attempt to digitize. These 117 files take up over 110 MB of space, meaning I found out right away about the 100MB quota. (Incidentally, devs, you're very open about the limit on the forum so why not just say it up front so that new users don't learn about it via error messages?)

I'd like to run some sort of batch command on these files with imagemagick or something similar but would like to know roughly how far down I can scale the images before they become unusable. Please let me know if there's some numbers already out there. In the meantime I'll keep testing on individual files to see if I can find the breaking point.

[engberg 89]

For cleanly scanned text, you'll get good results if a lower-case letter 'o' is 10-15 pixels high. Really huge letters aren't actually any easier for us to read ... in fact, it can occasionally make things worse due to extraneous detail that can throw us off.

It's a little hard to say an exact image size (e.g. 1280x960) that will work best, since it's more of an issue of the size and clarity of the text. You may find that a lot of cameraphones have a large number of pixels, but have lenses that don't focus well on close objects like business cards, which become highly blurry. This means you may get better results from a 640x480 picture of a distant billboard using an iPhone than you would get at max resolution for a business card an inch away from the lens.