5

I'm looking for software that searches through multiple images or a PDF for an smaller image that I give it.

The best example I can give is with this page from a stamp catalogue:

enter image description here

I would give the software a scanned stamp image, and it will find (all) candidates.
I know this is not easy, and a lot depends on image size and quality. Assume good quality scans/PDF and pictures of approx. equal size (again, looking for a stamp in a professional catalogue is a good example).

  • Windows
  • Free is always nice ;-)
  • Input image type can be anything common (BMP, JPG, PNG)
  • I would prefer searching though a (multiple page) PDF, but I can convert the PDF to images for each page if need be. Then the tool should check against a folder full of images.
| improve this question | |
  • You're asking for a 500000 $~ piece of software (if works well), just for you to know. I'll give you the fihsing rod, and let you go fish youself: If your pdf is composed of text with images, you just have to iterate over every image doing comparision with your stamp. Use some threshold to determine that two photos are identicall. There are also the common segmentation and matching techniques used in IA to recognize images. If your PDF if composed of full page images with the stamp inside the sa me image, you need an aditional step, which is detect that the rectangle (...) – DGoiko Nov 16 '18 at 11:21
  • A inside the big picture is a photo of a stamp, and then perform the stamp recognition steps only on that area. Study image segmentation and recognition, as long as image procesing using convolutional neural networks, it may give you a starting point. By the way, I'd recommend you to "convert" your big photo to the size of the comparing target to make things easier, however, if not all stamps are the same size, you'll need multiple conversions. I work on IA and I can tell you this is not a cheap code (if you want it to get higher than 90% accuracy and coverage. – DGoiko Nov 16 '18 at 11:23
  • iteration over images is quite simple, if you have the correct PF tools. Depending on your languaje there are plenty of free libraries (just using my memory, I can tell you PDFBox / Tika for Java or pyPdf2 / PDFMiner for Python, which are the most widely used programming languajes for IA prototyping right now. – DGoiko Nov 16 '18 at 11:27
  • TL; DR: The software you're asking for does not exists, it has to be done ad-hoc, and any serious company with serious results (higher than 90%) will charge you, at least, 500.000 dolars for what you're asking. The only way is to make it youself. If it doesnt involve languaje analyzing at all, you could contact IBM, although they overcharge shitty solutions because their name is famous, but they're terrible @ IA, better go for a local IA enterprise, preferably closely related to a university and with PhDs on IA on the payroll, or maybe Google has something for you, but it wont be cheap. – DGoiko Nov 16 '18 at 11:28
  • Related: softwarerecs.stackexchange.com/questions/21056/… – user416 Apr 26 '19 at 12:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy