0

I used hardware solutions like the Fujitsu iX500 ScanSnap scanner for on the fly OCR while scanning and I also used Foxit PhantomPDF to add text (OCR) to images in a PDF.

Both solutions resulted in not bad OCR but there are some flaws like false positives on non text or 99% of the text being missed on a page because there are 3-4 words in the corner of the page that are rotated 90 degrees in another direction as all the other text on the page.

At least PhantomPDF allow you to reviews what the software think might be false positive but you don't have much control.

What are the state of the art OCR solutions to either convert images to searchable PDFs or to add (hidden) text to existing PDFs only containing images (add search feature)?

2

If a programmatic solution will work, Syncfusion Tesseract OPX has a wrapper for the Tesseract engine that can be used to process the text in images within PDF documents and overlay them with searchable text.

//Load a PDF document
PdfLoadedDocument lDoc = new PdfLoadedDocument("Input.pdf");
//Set OCR language to process
processor.Settings.Language = Languages.English;
//Process OCR by providing the PDF document and Tesseract data
processor.PerformOCR(lDoc, @"Tessdata\");
//Save the OCR processed PDF document in the disk
lDoc.Save("Sample.pdf");
lDoc.Close(true);

Tesseract OPX is free, but Essential PDF is a licensed product. The whole product is also available for free through the community license if you qualify (less than 1 million USD in revenue).

Note: I work for Syncfusion.

  • It is worth noting that the accuracy of any OCR will depend on a large number of factors including, but not limited to, the language both which and selected correctly, the content highly technical or mathematically text often OCRs poorly, the font(s), clarity, alignment & cleanness of the images & the originals, layout & context. You may be able to improve this by training Tesseract - this looks helpful: pretius.com/…. – Steve Barnes May 26 '18 at 5:48
  • Davis, it would also be a good idea to make it clear that Syncfusion Tesseract OPX is a costly library to allow the development of such a program not a standalone program in its own right. – Steve Barnes May 26 '18 at 5:53
  • @SteveBarnes Thanks. I have added the details. – Davis Jebaraj May 30 '18 at 17:58
1

There's another way to OCR documents and images and save as searchable PDF besides using a programming library. The LEADTOOLS ePrint printer driver and converter has an OCR option. (Disclaimer: I am an employee of the vendor of this driver). To use it, you print the source document or image and select the OCR printer. You can save the printout as PDF or a number other document formats.

It has a "smart" setting that you can select to make it only OCR jobs that consist of images and don't already have text. It also supports saving PDF as "image over text" to keep the original document or image appearance, while making it searchable. If you'd like to try it, there's a time-limited free evaluation edition here.

Your Answer

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

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