I want to be able to find text on the following awesome image and where the text is located in it. This is not the first time I stumbled upon such a challenge and probably many people have similar questions from time to time as well. I think this stands as a good instance of the general issue.

There are many ways (1, 2, 3, 4 ...) to achieve this in a customized fashion but is there any OCR out there able to automagically transform this into a texted PDF? Or any other format for that mater. I'm citing PDF just because it is the format made to handle exactly this kind of stuff.

Looks like Cuneiform would be a perfect tool to do it manually, but I couldn't compile it on the mac and I wouldn't bother doing it for this instance, but I bet some relentless intern could use it to complement the OCR which would never be able to identify the images in it.

Here's a reduced sample of the image so we can better picture the relevance of the question:

awesome image

  • 3
    Ok, seriously, you don't need to put a massive image in your post when it has nothing to do with the question. – Wuffers Mar 23 '11 at 20:26
  • 5
    @Mark it has everything to do with the question. I want to find text on that image. It perfectly illustrates the urge. – cregox Mar 23 '11 at 20:49
  • It doesn't. Really. Your question applies to any image. What use would the question be to other people if it only applied to that image? – Wuffers Mar 23 '11 at 20:51
  • 2
    @Mark without the image the question might not be clear enough. I linked it not to display the image, but to give the source. You now know exactly what I meant because you looked at it before editing it out. But what would you picture the question would be if you haven't looked at it? I bet you wouldn't think it was that complicated. – cregox Mar 23 '11 at 20:54

Some almost-solutions:

Google Docs will OCR but not insert the text as a layer (I've not tried it out myself) http://googledocs.blogspot.com/2010/06/optical-character-recognition-ocr-in.html

Abbyy Finereader (http://www.abbyy.com/) will make it into a PDF with underlying text (I currently use it for this purpose), however it does cost (you could grab the trial version) and is Windows only

Evernote can OCR within images however I don't know if it will export to a PDF with underlying text

You could download a trial version of Acrobat Pro and use Document > OCR Text Recognition > Recognise Text Using OCR. I do have Acrobat Pro at work, but never used the OCR function so don't know how good it is. Again, Windows only (but you might well have Windows on a VM/Parallels/Virtualbox/Bootcamp)

  • I've tried google docs. It mostly sucked bad on that image. Just try it yourself! Evernote does take a while to OCR up, but it's freaking awesome. Too bad no PDF (or good way to see it anyway). I could only use it inside the program itself and amazingly enough I couldn't find there any way to "zoom in". And, as weird as it may sound, there is no acrobat trial for mac - but I'll give it a shot soon enough. – cregox Mar 24 '11 at 0:03
  • So far Evernote was the only solution. If anyone is interested, I also got to this stunning script, but I couldn't make it work because I have no root access at this work machine. – cregox Apr 4 '11 at 20:09

The image you link has such a low resolution that most OCR software will have a hard time with it, and you may end up with a pretty bad solution. Honestly, if you started transcribing it yourself now you'd probably save time over finding a few packages, trying them out, trying to get them to work, and ultimately having to fix every other entry due to misreads.

  • Adam, read the comments on the question. There are two links with same image in high resolution. Unless you mean even that has a low resolution - to which I would agree to a point. And I don't care all that much if this is never done. I probably would never bother to transcribe it myself. At least not this time, not with this one image... ;) – cregox Mar 23 '11 at 23:54
  • 1
    @Cawas - The linked "high resolution" image isn't high enough resolution. However, you have the source for the original (in the lower left corner) have you considered contacting them and requesting the original design file, which is likely to include text? – Adam Davis Mar 23 '11 at 23:58
  • Adam, not really. The source is also on the first link. But, again, can't care about it that much. My point here is the OCR challenge! – cregox Mar 24 '11 at 0:06

Not the answer you're looking for? Browse other questions tagged or ask your own question.