Argument 1: SCENARIO I submit a search via Google or Stack Exchange on some topic which I have recently been researching. I know I saw a good answer to my question somewhere - but I just can't remember where. Strangely neither Google, nor Stack Exchange - nor my textually indexed Safari browser history - shows any of the related words that I'm positive were in the context of the material that I had seen.

enter image description here

See the following question and answer which had ultimately been what I was looking for. However, this poster, whether justified, or not.. had posted (relevant and interesting) image content, and subsequently that information was in no way indexed or queryable.

My proposed solution is to run a simple OCR routine on posted images and include that metadata in the scope of the question.

Argument 2: This seems to be a technology / approach that is implemented elsewhere and that provides meaningful added value to users.

enter image description here

EDIT For those hooing haawing about the insane processing demands that this would require… I setup a quick test suite using google's open-source Tesseract on my Mac (brew install tesseract) which I encourage you to try at home ;-)

PNG=/tmp/test.png; TXT=/tmp/text; curl -L http://links.mrgray.com/ocrtest > $PNG SECONDS=0; /usr/local/bin/tesseract $PNG $TXT cat $TXT.txt; echo "that took approximately $SECONDS second(s)"

The Image.. enter image description here

The result…

Tesseract Open Source OCR Engine v3.01 with Leptonica
Certainly theology needs empirical facts and scientific theoretical
insights. The social scientists offer help. Yet they do not accomplish
what I must now attempt. My main question is where and how the
church must stand to be the witnssing church; that is, what must be
the relation between the culture that is the church (and the larger
Christian and biblical metaculture the church represents) and those
cultures the church indwells, evangelizes, serves? Answering will
require all the resources that Christian theology can bring to bear, and
not a little help from such as Berger and Bellah as well. Already they
have showed us, willy»nilly, that theology is required for the task: they
make such ample (and often skillful) use of it, themselves!

that took approximately 1 second(s)

Stack exchange is more than welcome to use my computer to handle any excess CPU-time required(if they want to pay for my cable modem, lol).

  • 7
    In all honesty, textual content should rather not be part of an image to begin with. We recently even corrected an issue where a user asked a question for which the text was largely contained in an image. I say that is better avoided altogether.
    – Bart
    Aug 4, 2012 at 16:18
  • I agree - and in an ideal world, posters would know not to "trap" vital content within an image.. Alas we live in an imperfect world - and the beauty of technology is that it can make up for some of our more dogged imperfections.
    – alex gray
    Aug 4, 2012 at 16:21
  • I imagine that would be quite a significant amount of work for what really should be an edge case, if it shouldn't be stopped altogether.
    – Bart
    Aug 4, 2012 at 16:22
  • 5
    You do realize that processing a single image on your dedicated home computer is vastly different than dealing with the sheer volume of images and posts involved in a site like this? This would involve a massive amount of effort for an edge case that we don't really want to encourage anyway. Aug 4, 2012 at 20:54
  • 1
    @BradLarson I guess that's a rhetorical question, but Yes? How many then? 10 per second? 100 per second? I doubt that many.. but whatever the amount it would not be more than a blip on the computational radar, overall. It's a single transaction, that could be scheduled off-peak, if that's really the objection.. and since when is requiring effort a reason not to do something.. if the result would provide more, and more accurate search results, help integrate and understanding user-contributed content, and generate no-hands metadata? Edge case? See response to dmckee.
    – alex gray
    Aug 4, 2012 at 21:25
  • 5
    @alexgray - Well, we're at 10.4 million total questions and answers on SO, although I can't find an easy way to query for the number of those that contain images. The Stack Exchange developers are already incredibly busy with adding features and helping the system to scale, and every feature they add takes time, is a potential source of bugs, and has to be maintained going forward. There is only a finite amount of developer time available, and as Bill says this is better solved by the community in that we should add the appropriate text ourselves instead of relying on images. Aug 4, 2012 at 21:38

2 Answers 2


I vote against. Textual content should be posted in text for exactly this reason, and way upwards of 99% of the time it is. Indexing text in images would be a lot of extra processing for very little benefit, and it would just encourage people to behave in a way we don't want.

  • 1
    Don't tell them - just DO it! The court system, police departments, jails, and halfway houses don't exist for the encouragement of criminals, but they do help with the problem…
    – alex gray
    Aug 4, 2012 at 18:37
  • @alexgray Bad analogy. The criminal justice system doesn't make it easier for someone to be a criminal either, it makes it harder. Aug 4, 2012 at 21:58
  • The criminal, in this case is not the user – everyone is allowed, rightfully to upload photos. The crime / victim here is that this data-focused ecosystem is yet to tap a rich, untapped source of info.. that is now only accessible only via direct, physical access. Enabling a way for the backend systems to provide greater detail about uploaded content is not going to cause people to all start rendering their posts as images and uploading them.. It will be the "cop" that stops that information from disappearing into a void.
    – alex gray
    Aug 4, 2012 at 22:29
  • 1
    @alexgray Overextending a bad analogy doesn't make it better. Aug 4, 2012 at 22:37
  • 1
    I never get all the negativity on meta - so here's another analogy.. :) There's a worldwide pandemic - that despite grave warnings - people thoughtlessly retransmit, suffer, and die from. Although incurable, there is a medicine that eliminates all symptoms. Alas, the zombied masses refuse treatment - as the disease makes them stupid - and aid workers are too scared/lazy to succeed with vaccinations. This idea - is that same medicine - administered like fluoride, via drinking water. Disease latent in us all, but humanity saved. Oh, and snarkiness isn't a reason it's not a good idea.
    – alex gray
    Aug 4, 2012 at 23:34

Seems like a lot of processing for a rare use case.

It be nice in a perfect world, but low, low priority.

  • I wouldn't consider photo posting to be an edge case use.. and with new sites popping up all the time, i.e. photography, graphic-design, etc… I would say it may be time to start to at least think about having this kind of feature.. even if it's not "user-facing"
    – alex gray
    Aug 4, 2012 at 18:43
  • 2
    Posting a image with text that represents important search terms that are not to be found in the accompanying text. Yeah, its a rare case. Aug 4, 2012 at 18:45
  • Not going to say you're wrong on something that's a: subjective and b: impossible to know for sure without "official" site stats.. but to see, I went to the apple SE site and opened the first question with a bounty and imagined what additional, relevant, and easily OCR'd info could have been gleaned. the answer. a lot.
    – alex gray
    Aug 4, 2012 at 18:50

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .