How to Remove Blank Pages from a Scanned PDF

A blank scanned page is rarely white: it has a grey background, dust specks and a streak from the roller. The detector renders every page, removes isolated specks, measures the share of dark pixels and counts the characters in the text layer. A page goes only when both checks call it blank. The sensitivity level decides how "dirty" a sheet still counts as clean, and pages with a lone signature or stamp are best protected with text recognition or checked by hand.

A stack of single-sided certificates went through the document feeder in duplex mode, and now every second page of the PDF is blank. Or a forty-page contract where half the sheets were printed on one side only now has twenty empty pages. Sending it looks careless, printing it wastes paper, and deleting pages one by one is slow and error-prone.

Remove blank pages finds those sheets automatically and strips them in a single pass. But a "blank page" means different things to a scanner and to a person, and whether the tool also throws away a page with a signature on it depends on exactly how the detector makes its decision. This guide covers how it works, what its settings control and where to double-check it.

Where blank pages in a PDF come from

The usual culprit is the scanner's duplex mode. The feeder captures both sides of every sheet, and if the back is clean, the file gets a blank page. For a stack of single-sided documents that is exactly half the file.

The second cause is separator sheets. Clean sheets go between documents so the combined scan can be split later, and nobody takes them out. The third is export from word processors: a section break or a stray page break at the end of a document produces a white page that was never in the text.

The fourth is worth remembering before you run the tool: blank pages put there on purpose. In books and manuals a new chapter starts on a right-hand, odd-numbered page, with an empty back side before it. In a document prepared for double-sided printing those sheets hold the layout together. Removing them shifts every chapter onto an even page. For print-ready files, delete pages by number rather than with the detector.

Why a blank scanned page is not white

Zoom in on a "blank" scanned page and it turns out not to be white. The paper has a grey tint, dust specks from the glass are scattered across the field, a thin streak from the feeder roller runs along the edge, and if the other side was printed, mirrored text shows through thin paper. To a program looking for pure white pixels, such a page is full of content.

So the detector does not look for white. It measures how much of the page is darker than the paper. Each page is rendered to an image and converted to greyscale. Then a median filter passes over it: every pixel is replaced by the median of its neighbours, so isolated dust specks and scanner noise disappear while solid marks, letters and lines survive. After that the tool counts the share of pixels darker than the paper threshold. On a truly empty scan that share is close to zero; on a page with a single paragraph of text it is noticeably higher.

The filter solves the main problem with scans: without it, a sheet sprinkled with fine debris would look full to the detector and stay in the file.

The second check: the text layer

The image is not the only source. If the PDF has a text layer, the tool also counts each page's characters, excluding spaces. This matters in two cases.

The first is documents from word processors. A page with the single line "Appendix 2" contains almost no ink, but it has text, so it stays. The second is the reverse: a page with a text layer that says "This page intentionally left blank" also stays, because it holds more characters than any sensitivity level allows. That is exactly the right behaviour: someone put that page there deliberately.

For a page to be removed, both checks must agree: ink below the threshold and characters below the threshold. One "filled" check is enough to keep the page. A scan with no recognised text has no characters at all, so the decision rests on the image alone.

What the three sensitivity levels control

The setting is called "Blank page sensitivity", and the level sets both thresholds at once: how much of the area may be covered by something dark, and how many characters of text are allowed for the page to still count as blank. The higher the level, the "dirtier" a sheet the detector is willing to call clean.

For orientation: at Medium a page counts as blank when less than one percent of its area is darker than the paper and the text layer holds no more than a few characters. Low is twice as strict, High twice as lenient. For comparison, a line of small text on an A4 sheet covers well under one percent of the area, so at High a scan with a single short line can go.

Hence the practical rule. Start at Medium. If grey blank sheets survive, raise the level. If something you needed is gone, lower it. Every run processes the original file from scratch, so trying different levels costs nothing.

The percentage threshold for manual tuning

Below the levels sits a percentage field, "Blankness threshold (%)". It is the same area share, given as a number: a threshold of 95 means a page counts as blank if at least 95 percent of its area is clean, in other words no more than five percent is dark. The higher the number, the stricter the check and the fewer pages get removed.

While the field is untouched it plays no part, and the threshold comes from the sensitivity level. As soon as a value is entered, it replaces the area threshold chosen by the level, while the character threshold still comes from the level. So reach for it only when the three levels did not fit: say, a very dark scan whose background covers several percent of the area and which even High refuses to call blank. Lower the threshold below the level's default, and those sheets go.

The opposite happens too: if the tool refuses to run because it considers every page blank and the threshold is already at its maximum, the error message will suggest lowering the sensitivity. The threshold itself cannot help at that point.

"Keep first page"

A separate toggle protects the first page even when the detector judged it blank. It exists for covers and title sheets: in scanned folders the first page is often an almost clean sheet with one caption or stamp, which is easy to lose at High. With it on, the first page stays and the rest are checked as usual.

The toggle does not extend to other pages. If you need to keep a clean sheet in the middle of the document, say a separator before an appendix, you will have to put it back afterwards, or remove blank pages by number instead.

Step by step

1. Open Remove blank pages and upload one PDF. The tool processes one file per run: combine a batch of scans into a single document first. 2. Leave sensitivity at Medium for the first pass. If the first page is a title sheet, switch on "Keep first page". 3. Start processing. Pages are checked one by one, so on a long scan the progress indicator moves evenly instead of stalling and then jumping to the end. 4. Download the result and check it as described below before deleting the original.

If the file has an open password, the tool will not accept it: remove the password with Unlock PDF first, then remove the pages. Protection that only forbids printing or copying does not get in the way. A damaged file that will not open should go through Repair PDF first.

How to check the result

The tool does not show a list of removed pages, but checking its work takes a minute.

Start with the page count. For a duplex scan of single-sided documents it should drop by exactly half: forty becomes twenty. Nineteen means one filled page went out with the blanks, and you need to find it. Twenty-three means the detector missed three blank sheets, and the level should go up.

Then leaf through the file where blank pages definitely were: the backs of certificates, the boundaries between documents, the end of the file. Neighbouring pages should run on without gaps. Look especially carefully at sheets with little content: a signature, a stamp, a single line with an appendix number, a handwritten note. Those are the ones closest to the threshold.

Keep the original until you are sure of the result. The tool never modifies it: removal happens in a new file.

Pages the detector may remove by mistake

Every automatic detector has a risk zone, and this one's is clear: scanned pages with little ink and no text layer.

• A sheet with a single signature or initials. The signature covers a fraction of a percent of the area, and the scan has no text layer.
• A page with a seal or stamp and no text around it.
• The last page of a contract carrying only the line "Attachments: 3 sheets" in small type.
• A form with a single field filled in.
• Faint pencil notes on an almost empty sheet.

At Low such pages almost always survive, at Medium they usually do too, at High they can go. If your document has sheets like these, run at Low or give the detector a second anchor in advance.

That second anchor is the text layer. If you run the scan through OCR PDF before removing blank pages, recognised text appears on every page that has anything readable: an appendix number, a stamp with an organisation's name, a line in small type. To the detector such a page stops being blank regardless of how much ink the image holds. Truly empty sheets get no text and will still be removed. Recognition will not save a signature without a single letter in it, so check those pages by eye.

What else changes in the file

Removing pages is more than hiding them from the viewer. The tool assembles a new document from the remaining pages, repoints internal links and bookmarks to the new numbers and drops objects that nothing references any more. The content of removed pages does not stay in the file. That matters if a "blank" page was actually a show-through back side carrying someone else's data: after processing it is gone from the file.

Page numbers printed on the sheets are not recalculated: they are just text in an image. If you need continuous numbering after the cleanup, add it again with a separate tool.

In what order to process a batch of scans

Removing blank pages is rarely the only operation on a scan. The file usually also needs assembling from parts, rotating, recognising and compressing, and the order of those steps affects the result.

Assemble the file before the cleanup, not after. Merge the documents first and the blank sheets leave all the parts in one run, with no need to repeat the settings for every file. Rotating upside-down pages can be done at any point: the detector looks at the share of ink, not at the orientation.

Put text recognition before the cleanup if the document has pages with little content, and after it if there are none. In the first case the text layer will protect signature sheets and stamps; in the second, recognition runs faster because blank pages no longer eat up time.

Leave compression for last. It re-encodes the page images, and every extra page costs time and space at that moment. Besides, heavy compression shifts the tint of the grey background, and a threshold that suited the original scan may behave differently. The detector should see the file as the scanner produced it.

For an ordinary stack of single-sided documents the sequence is: combine the parts into one file, remove blank pages, rotate if needed, recognise text, compress before sending. For a folder with signature sheets and forms, recognition moves to second place, before the cleanup.

When deleting by number beats the detector

Automation suits scans where the blank pages are all equally blank and there are many of them. In three cases the manual route is more reliable.

The first is a file meant for double-sided printing, where the empty backs are there on purpose. The second is a document with several nearly empty but necessary pages: signature sheets, forms with a single field. The third is when there are only two or three blank pages and you know where they are. Then open Remove pages and enter the numbers, or rearrange and drop the extras by eye in Organize PDF, where every page is visible before it disappears.

And if you have not scanned yet, it is easier not to create the problem: for single-sided documents turn off duplex mode in the scanner settings, and the empty backs never reach the file. When you scan with a phone, Scan to PDF assembles only the shots you took, so it contains no blank pages by construction.