OCR pdf perfomance

[mirkop]

Hi,
I have installed the lastest version of GdPicture 8.4.
I read the text from pdf file, but it's too slow.
I have some files (total 85 MB), and for read the text of all pdf the process during 2 hours.
Foreach file use this code:

Code: Select all

GdPicturePDF oPDF = new GdPicturePDF();
                    if (oPDF.LoadFromFile(path, false) == GdPictureStatus.OK)
                    {
                       
                        int dimCount= oPDF.GetPageCount();

                        for (int i = 1; i <= dimCount; i++)
                        {
                            if (i > 1)
                            {
                                Debug("SELEZIONE PAGINA " + i);
                                oPDF.SelectPage(i);
                            }
                            Debug("RenderPageToGdPictureImage");
                            m_ImageID = oPDF.RenderPageToGdPictureImage(200, true);

                            Debug("OCRTesseractReinit");
                            oGdPictureImaging.OCRTesseractReinit();

                            Debug("OCRTesseractDoOCR");
                            s += oGdPictureImaging.OCRTesseractDoOCR(m_ImageID, "ita", _dirOCR, "");
                            if (oGdPictureImaging.GetStat() != GdPictureStatus.OK)
                                Debug("[" + path + "] Error on page " + i + ": " + oGdPictureImaging.GetStat().ToString());
                            Debug("OCRTesseractClear");
                            oGdPictureImaging.OCRTesseractClear();
                        }
                        oPDF.CloseDocument();
                    }

How can i improve the performance?

Thank you

Mirko

[Loïc]

Hi Mirkop,

You can decrease the PDF dpi rendering or OCR your document in a multi-threaded application.

Steps are:

- Split the document (one new doc per page)
- in X threads create 1 PDF/OCR per page
- At the end, compose a new PDF by merging all document produced

We should deliver such demo application within the next release.

Kind regards,

Loîc

[mirkop]

Hi Loic,

thank you for your reply.
I don't need to create a new file, but only exctract the ocr text.

Is there a way to improve the performace?

Mirko

[mirkop]

any suggestion ?

Mirko

[Loïc]

Mirko I already provided 2 suggestions:

1- Decrease your PDF rendering resolution.

2- use multiple threads splitting your input document.

I don't see other possible way. I will try to provide a demo for (2) as soon as I can.