OCR accuracy V6 vs V7

[DWC121]

Greetings - I'm using two versions of PDF-XChange Editor. The one on my desktop is V7.0 build 326.0. My portable copy is V6.0 build 317.1 . My portable copy is older but the OCR functionality is way more accurate. An example of both are attached. Both pdfs were created from the same bmp file.

I'm almost sure I have all the settings the same for both (Lang: English, Accuracy: Medium, Output type: Preserve Original Content and Add Text Layer, Quality: 300). Is there another setting in my V7 I need to adjust to increase the accuracy? Is there a "hidden" file somewhere (maybe in my V6) that I can import into V7 to increase the accuracy?

A few weeks ago I tried darkening and sharpening another bmp, but the result in V7 was still sub-par compared to V6.

Thanks - David

I'm using Windows 7 Home Premium edition

[Timur Born]

Is "medium" OCR quality fixed already? Quite some time ago it was written that the whole OCR engine would get an overhaul, but I never heard about it again.

[Willy Van Nuffel]

Hello,

Tracker Software introduced a renewed OCR feature a few releases ago (starting from 7.0.325):
= "Enhance Scanned Pages" - in the Document-menu - Classic Toolbars / in the Convert-tab - Ribbon UI.

I suppose this one will replace the current "OCR Page(s)..." feature over time.

See:
viewtopic.php?f=65&t=30794
viewtopic.php?f=62&t=30797

Regards.

[TrackerSupp-Daniel]

Hello DWC121,

Thank you for those comparison files. Could I ask that you try using the new "Enhanced Scanned Pages" function in the latest build and tell us if the results are more desirable?

Thanks for the forum references Willy
Timur, Willy is correct in that the "Enhanced Scanned Pages" option is the OCR overhaul that was written about, Its development is still ongoing, as such we do of course still offer both options for users.

[DWC121]

Daniel,

I tried the new "Enhanced Scanned Pages" function and the results are much better. The original image is on the light side (attached as PDF TEST.jpg). Note that when I create a pdf I import a bmp file; not a jpg file.

The "Enhanced Scanned Pages" function has several options. In my case I found the following settings made no difference:
"Color/Grayscale" = JPG, default quality
"Color/Grayscale" = JPG, "Small size"
"Color/Grayscale" = JPG, "High Quality"
"Color/Grayscale" = JPG2000, default quality
"Color/Grayscale" = JPG2000, "Small size"
"Color/Grayscale" = JPG2000, "High Quality"
Even though my image is monochrome I left that setting as JBIG2.
For the above, I left the "Filter" options at their default settings (On, off, low, low)

I then tried the above JPG combinations with "Filters - Text Sharpening" at Medium and again at High. Both gave satisfactory results. The OCR were minimally different. These files are attached. The High setting made the text much darker (as expected). The Medium setting kept the tonal color of the black text very close to the original. I prefer that since the pdf might be printed later. I need the pdf to appear close to the original bmp file.

It is great there are many options for applying OCR. I just hope people don't get set in their ways and use one particular set of options... and then find out later after applying OCR to hundreds of documents "I should have set another option differently".

David

[Tracker Supp-Stefan]

Hello David,

Glad to hear you are now happy with the results!

And many thanks for sharing your findings and samples!
I am sure they will be useful to others as well!

Cheers,
Stefan