[SOLVED] [GhostScript] Any way to shrink scanned page further?

littlebigman · 02-08-2022, 07:44 AM

Hello,

This is a question for GhostScript experts.

Is it possible to seriously shrink an 800-page PDF of a scanned book + OCRed text layer from its original ~100MB? FWIW, the whole book is in black and white.

The following test on a single page only reduces the size by ~15% (100kB → 85kB):

Code:

gs -sDEVICE=pdfwrite -dCompatibilityLevel=1.4 -dPDFSETTINGS=/screen -dNOPAUSE -dQUIET -dBATCH -sOutputFile=single.output.pdf single.input.pdf

Thank you.

business_kid · 02-08-2022, 01:32 PM

I've done a bit of that in my time, and the thing leaping out at me is: Get out of the pdf format if you possibly can.

Furthermore, if you have and need an OCR'ed layer for text, I presume it's because the book isn't very readable in it's present format. Now I suppose you were up to at least 400dpi for OCR, but that work is done, so why not reduce the dpi of the image now? They don't need such a great image of the book if you have text overlaid. You seem good with gs, but I would use console tools - ffmpeg or something.

littlebigman · 02-08-2022, 02:59 PM

The text layer is to make the text copy/pastable.

I guess there's no way to make it much smaller with no access to the source pictures.

Thank you.

business_kid · 02-09-2022, 04:36 AM

You should be able to copy a pdf with no text layer, afaik. PDFs also reduce the resolution of pictures automatically in making the pdf. If you put a 1200dpi picture into a pdf, the resolution will be reduced.If you go back to your gs man page, you can probably set how much it's reduced by. Make sure to keep a high resolution copy but the sizes I've seen are 600dpi(highest) and 300dpi (normal). You could probably set your chosen resolution.

littlebigman · 02-09-2022, 08:50 AM

Thanks.

I have another idea: Since there's a text layer that was created by the OCR, can I just remove the "picture" layer, and see if the text is in a readable shape? That would make for a much smaller PDF.

---
Edit: Easy enough :-p

Code:

gs -sDEVICE=txtwrite -o output.txt input.pdf

As expected, it's raw text, so some post-editing is required to get something that looks close to the original.

teckk · 02-09-2022, 09:31 AM

Let me see, cert.pdf is already a small 17 page text .pdf, no images.

Code:

du -c cert.pdf
212     cert.pdf
212     total

gs -dNOPAUSE -dBATCH -sDEVICE=pdfwrite -dCompatibilityLevel=1.4 -dPDFSETTINGS=/ebook -sOutputFile=Out.pdf cert.pdf
...
Page 13
Page 14
Page 15
Page 16
Page 17

du -c Out.pdf
240     Out.pdf
240     total

gs -dNOPAUSE -dBATCH -sDEVICE=pdfwrite -dCompatibilityLevel=1.4 -dPDFSETTINGS=/screen -sOutputFile=Out.pdf cert.pdf

du -c Out.pdf
240     Out.pdf
240     total

gs -dNOPAUSE -dBATCH -sDEVICE=pdfwrite -dCompatibilityLevel=1.4 -dPDFSETTINGS=/printer -sOutputFile=Out.pdf cert.pdf

du -c Out.pdf
244     Out.pdf
244     total

gs -dNOPAUSE -dBATCH -sDEVICE=pdfwrite -dCompatibilityLevel=1.4 -dPDFSETTINGS=/prepress -sOutputFile=Out.pdf cert.pdf

du -c Out.pdf
240     Out.pdf
240     total

So for that .pdf, which is already tight and small, that's about as good as you can get with -dPDFSETTINGS

Make the page size smaller.

Code:

gs -dNOPAUSE -dBATCH -sDEVICE=pdfwrite -dCompatibilityLevel=1.4 -sPAPERSIZE=a6 -dFIXEDMEDIA -dPDFFitPage -sOutputFile=Out.pdf cert.pdf

du -c Out.pdf
244     Out.pdf
244     total

Change the dpi

Code:

gs -dNOPAUSE -dBATCH -sDEVICE=pdfwrite -dCompatibilityLevel=1.4 -r200 -sOutputFile=Out.pdf cert.pdf

du -c Out.pdf
236     Out.pdf
236     total

Ok Lets make a poor quality pdf with a small page size

Code:

gs -dNOPAUSE -dBATCH -sDEVICE=pdfwrite -dCompatibilityLevel=1.4 -r100 -sPAPERSIZE=a10 -dFIXEDMEDIA -dPDFFitPage -sOutputFile=Out.pdf cert.pdf

du -c Out.pdf
224     Out.pdf
224     total

It was already a real small pdf to start with. Probably get more reduction in size with a busy pdf.