← All posts

Delete the blank pages a duplex scan left behind

19 August 2026

You put a stack of paper through the document feeder, ticked double-sided because some of the sheets were, and got back a 200-page PDF where half the pages are empty. Page 1 is a letter, page 2 is nothing, page 3 is the next letter, page 4 is nothing again.

Nobody wants to click through 200 thumbnails to write down which ones to remove. And the moment you try, you find the other half of the problem: the empty pages are not empty. They are photographs of the back of a sheet of paper, and every tool that claims to "find blank pages" is really guessing at a threshold.

There are two versions of this job and they are not the same amount of work. If every second page is blank, you are one step from finished. If the stack was mixed, you have to find them first.

Why the scanner made them

A duplex feeder photographs both sides of every sheet as it passes the sensors. It has no idea whether the second side was printed on — it is a camera, not a reader. So a single-sided sheet produces two images: your letter, and a picture of the blank back.

That is why the pattern is so regular when the whole stack was single-sided. Sheet one gives pages 1 and 2, sheet two gives pages 3 and 4, and the blanks land on every even number. It is also why the pattern collapses the moment one sheet in the pile was printed on both sides: from that sheet onwards the parity flips, and the blanks are on the odd numbers instead.

Why a blank scan is not blank

This is the part every other page on this subject skips, and it is the reason the job is harder than it sounds.

A born-digital blank page — one that came out of Word — genuinely contains nothing. There is no drawing on it at all, so a program can say "this page is empty" with certainty.

A scanned blank page contains a full-size image of white paper, and that image carries:

• Show-through. The scanner lamp shines through 80 gsm paper, so the printing on the front turns up as a faint mirror image on the back. On a dark page — a photograph, a solid heading — it is very visible.
• Dust, hole punches, staple shadows and the grey band down the edge where the sheet did not sit flat against the lid.
• Sensor noise. Every scanner has a noise floor, and JPEG compression turns it into faint mottling.

So "is this page blank" is never a yes-or-no question about a scan. It is a question about ink coverage: what percentage of the pixels are dark once you have thrown away everything below a threshold. A blank back is typically under 0.1%. A page carrying a single line of text is around 1%. That gap is what every script in every forum thread is measuring.

The regular case: every second page

If the whole stack was single-sided, do not list anything. Open Delete PDF pages and write even in the pages box. It removes pages 2, 4, 6 and so on in one pass, and tells you how many went and how many are left — check that 200 became 100 before you do anything else.

1. Look at pages 1 to 6 first. In any reader, at a thumbnail zoom. You are confirming that the blanks are on the even numbers and not the odd ones.
2. If they are on the odd numbers, write odd instead. That happens when the first thing in the feeder was a cover sheet, or when the scanner was set to place the back first.
3. Delete them with even or odd, and download the result.
4. Mind the last sheet. If the final sheet in the stack was printed on both sides, its back is real content sitting on an even number, and even will take it with everything else. The patterns apply to the whole document and cannot be limited to a range, so pull that page out of the original first with Extract PDF pages, then join it back on with Merge PDF afterwards.

The irregular case: find them, then list them

A mixed stack has no pattern, so somebody has to identify the blanks. Three ways, fastest first.

Sort the pages by file size. A JPEG of a blank sheet compresses to a fraction of the size of a page with printing on it. Run the file through PDF to JPG at 72 dpi, open the zip, and sort by size. The blanks are the smallest files, in a clump, and their numbers are in their filenames. This takes two minutes on a 200-page scan and is far quicker than looking at thumbnails.

Use a thumbnail grid. Any reader will show 4-up or 8-up. On a mixed stack this is fine for 40 pages and miserable for 400.

Check the text layer, if the file has been through OCR. Our PDF inspector lists every page with the number of characters drawn on it, and runs entirely in your browser. On an OCR'd scan, the pages showing zero or a handful of characters are your blanks. On a scan that has never been OCR'd, every page shows zero and the report tells you nothing — that is the honest limit of the method.

Once you have the numbers, put them into Delete PDF pages as a list: 6,14,15,22,40-42. If more pages are going than staying, invert the job: name the pages you want to keep in Organize PDF pages instead, or pull a run out with Extract PDF pages. Same result, less typing.

What our tool will not do

It does not detect blank pages. even and odd are patterns, not detectors. Nothing in the delete step looks at what is drawn on a page — it takes the numbers you give it and rebuilds the document from what is left. Anything less regular than every-second-page has to be listed by number, and finding those numbers is the section above.

It works on one file at a time, and the numbers are the file's own, counted from 1 — not the numbers printed in the footer of the paper.

The result is a rebuilt document. Pages you keep are copied into a fresh PDF, so bookmarks and form field definitions do not survive. On a scan there are none of either, so this costs you nothing here; on a born-digital document it matters, and it is worth knowing before you run it on a report.

What does detect them

We would rather you finish the job than use our tools. If this is a weekly task, or the stack runs into the thousands, do it properly:

• ImageMagick, ten lines of shell. Render the pages, threshold, and print the ink coverage of each one:

  pdftoppm -r 50 -gray -png in.pdf page
  for f in page-*.png; do
    echo "$(magick "$f" -shave 40x40 -threshold 60% -format '%[fx:100*(1-mean)]' info:) $f"
  done | sort -n
  

  The -shave throws away the outer edge, which is where the lid shadow lives and where a naive check goes wrong. Anything under about 0.1 is blank; look at the two or three either side of your cut-off before you commit to it.

• qpdf to do the removing, once you have the list: qpdf in.pdf --pages . 1,3,5-7 -- out.pdf names the pages to keep. pdftk in.pdf cat 1 3 5-7 output out.pdf is the same idea if you still have pdftk.

• ScanTailor Advanced is built for processing large scans. Its content-selection step draws a box around whatever is actually on each page, so a blank shows up as a page with no content box — you can see them all at a glance and drop them there.

• Acrobat Pro has a blank-page fixup in Preflight with an adjustable threshold, and an Action to run it over a folder. Worth knowing about if your employer already pays for it.

Stopping it happening again

Your scanner probably has the setting already: "skip blank page", "blank page removal" or "blank page detection", usually with a sensitivity slider, in the driver rather than in the software you scan from.

Turn it on and turn the sensitivity down, because its failure mode is the expensive one. A high setting throws away a page that carries one line — a signature page, a compliments slip, a form with three boxes ticked — and it does it silently, so you find out weeks later. Compare the page count of the PDF against the number of sides you fed in, every time, until you trust the setting.

Afterwards

Halving the page count halves the file, and the deleted pages take their images with them: the kept pages are copied into a new document, so nothing from the blanks is left behind as unreferenced data. If it is still too large for email, Compress PDF re-encodes the scanned images — expect a big saving on a colour scan and very little on one that was already greyscale. While you are there, a sheet-fed scan is usually a degree or two crooked, and Deskew PDF will straighten it.

Two neighbouring jobs, in case the one you actually have is one of them. If the fronts and backs came out as two separate files because the scanner is single-sided, you do not want to delete anything — you want to interleave them back into one document. And if what you need at the end is each surviving page as its own file, that is splitting into single pages, which is a different tool and has its own trap about file naming.