Wrong PDF when using OCR_MakeSearchable

baumunk · Post by **baumunk** » Fri Nov 13, 2020 9:41 am

We test your product.
With your example application OCRDemoCsharp

Pdf contains images and text. After OCR, text in the image is searchable but text is removed.

Orignal PDF:

After OCR:

What am I doing wrong?
We are using DEMO.

If I use the same function in PDF-XChange Editor it is correct.

Code:

Code: Select all

hResult = PDFXOCR.PDFXOCR_Funcs.OCR_SetCallback(pdf, thecallback, 0);

				hResult = PDFXOCR.PDFXOCR_Funcs.OCR_LoadW(pdf, m_SourceFilename);
				if (PDFXOCR.PDFXOCR_Funcs.IS_DS_FAILED(hResult))
				{
					MessageBox.Show("Error loading file: \n" + m_SourceFilename, "OCR Library Error");
					break;
				};


				PDFXOCR.PDFXOCR_Funcs.PXO_Options Options = new PDFXOCR.PDFXOCR_Funcs.PXO_Options();
				Options.blacklist = "";
				Options.whitelist = "";
				Options.raster_dpi = 300;
                Options.ImageFlags = (uint) PDFXOCR.PDFXOCR_Funcs.OCR_ImageProcessingFlags.OCR_Image_NoRotate;
				Options.DataPath = m_Datapath;
				Options.lang = m_Language;
				Options.RegionMode = PDFXOCR.PDFXOCR_Funcs.OCR_RegionMode.OCR_Auto;
				Options.SecondLanguage = 0;

				IntPtr pxoPagelist = IntPtr.Zero; // null pointer passed to OCR_MakeSearchable() will result in all pages being OCRd.

           
                hResult = PDFXOCR.PDFXOCR_Funcs.OCR_MakeSearchable(pdf, ref Options, pxoPagelist);

Regards.

Post by **Paul - Tracker Supp** » Mon Dec 07, 2020 8:57 pm

Hi baumunk,

I did speak to one of the dev team about this. He told me the issue is reproduced and to make a formal support request ticket around this. While intended for internal use only, if you refer to RT#5394: Wrong PDF when using OCR_MakeSearchable in correspondence with us then any support staff member can get yo a status update on the ticket.

Back to the issue itself, you should see a post here with some suggestions withing 24 hours.

YouTube · Post by **Sasha - Tracker Dev Team** » Tue Dec 08, 2020 10:42 am

Hello baumunk,

We've inspected this behavior and it seems that it's a bug. Part of the content is not being rendered and the SDK does not know about it.
But, as a temporary fix, you can try using the additional flags: OCR_Content_Original and OCR_Image_SuppressOutput. Then the text that is being lost in the rendering process will be left as is and the text for search will be put on top of it.

Cheers,
Alex

baumunk · Post by **baumunk** » Fri Feb 12, 2021 11:55 am

Hello Alex,

Thanks for the tip. This helped.
OCR_Content_Original = 0x0040
but was only in description not in PDFOCR_Funcs.cs

With kind regards
Ernest Baumunk

YouTube · Post by **Sasha - Tracker Dev Team** » Fri Feb 12, 2021 12:40 pm

baumunk · Post by **baumunk** » Tue Feb 16, 2021 10:28 am

I have another problem.

After OCR I need to convert file to PDF/A-1B.
Since we have PDF-XChange PRO SDK license. After OCR new file is printed to XChange PDF driver.

Although after that PDF is a valid PDF/A-1B.
But the OCR recognition is then no longer present.

Do you have a solution for this available?

On my other tiket: https://forum.tracker-software.com/viewtopic.php?f=43&t=35816

unfortunately nobody answers.

With kind regards
Ernest Baumunk

Post by **Tracker Supp-Stefan** » Fri Feb 19, 2021 11:44 am

Hello baumunk,

It seems like the PDF/A-1B part of your problem is being handled in that other topic - so shall we consider this one resolved and continue the discussion in the other one?

Kind regards,
Stefan