"Create PDF from Images": "Deskew" and "Detect skew of image"

[Jensen Head]

In the "Create PDF from Images" tool settings, in the "Image Post-Processing Options" tab, there is a "Deskew" checkbox. In addition, in the same tab in the "OCR Image (Enhanced)" window there is a "Detect skew of image" setting.

2021-11-03_14-05-26.png

Am I correct in understanding that the first setting automatically corrects a slight skew of the entire page (within ± 45 degrees), while the second leaves the rotation of the imported page intact, and only helps the OCR engine to detect text lines located at arbitrary angles?

[Radi - PDF-XChange]

Hello Jensen Head,

Both options do essentially the same - they detect if the image is skewed and straighten it. There are two separate options for this because you may want to deskew the image without running OCR. Also, the second deskew (the one from the OCR settings) is run by the OCR engine.

Regards,
Radi

[Jensen Head]

If both options are selected, couldn't it happen that the image will be rotated twice, that is, re-encoded twice with a corresponding deterioration in quality? If the user is going to perform OCR anyway, which method would you recommend as more accurate/correct/works with fewer false positives?

[Dimitar - PDF-XChange]

Hello Jensen Head,

I would recommend using the dedicated OCR tool first as it has much more options to choose frоm.

image.png

Regards.

[Jensen Head]

I would recommend using the dedicated OCR tool

I didn't quite understand your advice. In your screenshot, you have two skew correction methods checked simultaneously: "Detect skew of page content" and "Fix content skew and incorrect page rotation." My last question was about the side effects of using these two methods simultaneously (OCR-based and non-OCR-based).

[Dimitar - PDF-XChange]

Hello Jensen Head,

which method would you recommend as more accurate/correct/works with fewer false positives?

The reason I suggested using the OCR tool directly is precisely because it tends to produce fewer false positives as it offers more control over the process.

That said, both options serve different purposes, so using them simultaneously will not lead to unexpected results.


Best regards.