Scan a document from a photo: a straight A4 or Letter page
Take a photo of a page, upload it and click its four corners. You get a straightened A4 or Letter page, in colour, grey or black and white, to download as a PNG or a PDF. The photo is processed in your browser.
Opens without an account 3 times a day. A free account removes the limit.
- 4 cornersyou click them, they are not detected
- A4 and LetterPDF at the paper size
- 100 to 300 DPIfour steps: 100, 150, 200, 300
- 3 kinds of outputcolour, grey and black and white
Scan a document from a photo: what it is
This document scanner online is a free tool from EduFacturing for when there is no scanner at hand but there is a photo of the sheet. You click the four corners, and the tool works out the distortion and builds the straightened page pixel by pixel.
The output is colour, grey or black and white; in black and white a local Sauvola threshold cleans the background, so a shadow from your hand does not blacken the whole page.
Photo of a page to PDF
Straightens by four corners
You click the corners in order: top-left, top-right, bottom-right, bottom-left, and can drag any point. An outline with an area under 400 px² or one that is not convex is rejected with an explanation.
Gives a page at true size
A4 is 210 × 297 mm and Letter is 215.9 × 279.4 mm. Pixels depend on the DPI you choose, 100, 150, 200 or 300: A4 at 150 DPI is 1240 × 1754 px.
Colour, grey or black and white
Colour keeps the photo, grey turns it into grey levels, and black and white applies a Sauvola threshold, which also cleans the background, with a window from 7 to 41 px and a strength k from 0.10 to 0.60.
Downloads PNG and PDF
The PDF is one page at the paper size. Measured corners keeps the proportion of a postcard, a note or a ticket: the page is as large as the image, the long side is capped at 3000 px, and 1 px is 1 pt.
Scanning without a scanner
Scanning a document without a scanner helps when there is a photo of the sheet but no scanner at hand.
A notebook page to PDF
A page from a notebook or textbook becomes a clean A4 sheet for a PDF or for printing.
Documents for the archive
A contract, receipt or invoice photographed on a table becomes a PDF at the paper size, and the local threshold keeps the text readable in uneven light.
Before text recognition
The straightened page can be uploaded to a text recognition tool instead of the photo taken at an angle.
A sheet that is not A4
A postcard, note or ticket keeps its proportion in Measured corners.
How the photo is straightened
A photo of a flat sheet taken at an angle is distorted in a predictable way, described by a 3 × 3 matrix called a homography. Four corners are exactly the data needed to find it.
Four corners, eight equations
Each corner gives one equation for x and one for y. The homography has eight unknowns because the ninth element is 1.
Normalisation
The points are shifted to centre on (0, 0) and scaled to an average distance of √2 (Hartley's normalised DLT, 1997). The 8 × 8 system is solved by Gaussian elimination with partial pivoting.
Straightening back to front
For each pixel of the page the inverse homography finds a point in the photo, and the colour comes from bilinear interpolation of four pixels. That leaves no holes. Outside the photo it is filled with white.
Sauvola threshold
For each pixel the mean and standard deviation are calculated in a window around it. The sums come from integral images, so the time does not depend on the window size.
| Applies to | Formula | Meaning |
|---|---|---|
| Sauvola threshold | T = m·(1 + k·(s / R − 1)) | m is the window mean, s is the standard deviation, R = 128, k defaults to 0.34; below T is ink |
| Grey levels | grey = 0.299·r + 0.587·g + 0.114·b | the BT.601 weights for the red, green and blue channels |
| Size in pixels | px = mm / 25.4 · DPI | how many pixels the page has at the chosen DPI |
| PDF page size | pt = mm · 72 / 25.4 | A4 is 595.3 × 841.9 pt |
What is different here
The corners are yours
There is no automatic edge detection, so there is no guess that quietly skews the page.
The photo stays in the browser
The straightening is solved in the tool's own code, with no computer vision library; the pdf-lib library only assembles the PDF file.
The tool does not find the corners itself, does not recognise text (OCR), does not process several pages at once and does not remove handwritten notes or stains. It does not correct brightness: a photo that is too dark is better retaken. The whole sheet has to be in the frame.
A4 and Letter come out in portrait orientation; for a landscape sheet use Measured corners. Without a reference in the frame (a ruler, a coin) the true size in mm cannot be known, so that mode gives a file in pixels, not for printing to scale. With a large photo and a high DPI the page may briefly stop responding.
Questions about scanning with a phone
How to scan a document with your phone?
Photograph the whole sheet, upload the photo, click its four corners and download the straightened page as a PNG or a PDF. A photo that is too dark is better retaken.
How do I make a photo look like a scanned document?
Click the four corners of the page: the tool straightens the perspective. The corners are manual because a wrongly detected one gives a page that looks straightened but is slightly skewed.
Can I print the scan at true size?
Yes, with A4 and Letter. The PDF page has the paper size and the image fills it completely. With Measured corners there is no physical size: 1 px is 1 pt.
Is the photo sent to a server?
No. Reading, straightening, converting to black and white and PDF assembly happen in the browser, and the tool's code has no request that sends the file.
What does the strength k do in black and white?
A larger k removes more of the background. It runs from 0.10 to 0.60 and defaults to 0.34, while the window runs from 7 to 41 px and defaults to 19.
Is it free?
Yes. Without an account it opens 3 times a day; a free account removes the limit.
The document scanner is free. Pro opens the rest
Pro costs 15 EUR a month and opens 45 Pro tools.
Related tools
Image to text (Bulgarian OCR)Extract text from an image, a scan or the first page of a PDF online, free. Reads Bulgarian (Cyrillic) and English, and the image is never uploaded.
PDF tools: merge, split, shrinkMerge PDF without uploading: join files, split pages, shrink, number, add a watermark, lock and sign. Your file stays in the browser. Pro, 15 EUR a month.
DOCX to Markdown, HTML and PDFDOCX to Markdown online: a Word or LibreOffice document becomes Markdown, HTML, plain text or PDF. Free, 3 times a day without an account, no limit with one.
Image converter and resizerFree online image converter: JPG to PNG, WebP to JPG, HEIC to JPG. Reduce image size to a target in KB, crop and strip GPS. Runs in your browser, no upload.
Scan your photo
Opens without an account 3 times a day. A free account removes the limit.