How to Turn a Photo of a Document Into a Clean, Straight Scan
Photographing a document instead of scanning it is quicker and the result looks wrong: the page is a trapezium, the lines run downhill, and the edges are cut off or surrounded by desk. The fix is not rotation. It is perspective correction, and the Document Scanner does it from four corner points.
Why a photo of a page is never square
A scanner holds the sensor parallel to the page at a fixed distance, so every part of the paper is the same distance away and the result is a perfect rectangle. A camera held in your hand is at an angle, so the near edge of the page is closer than the far edge and therefore appears wider. A rectangle photographed at an angle is a trapezium - that is not distortion, it is geometry.
Rotating cannot fix it, because the two edges are genuinely different lengths in the image. What is needed is a transformation that maps the four corners of the trapezium onto the four corners of a rectangle, stretching the far edge and compressing the near one. That is what you are telling the tool when you drag the four points onto the corners of the page.
How to photograph it so the correction works well
Get the whole page in frame with a little margin, so all four corners are visible - the correction needs them. Shoot from as close to directly above as you can manage; a shallow angle means the far edge is being stretched a long way and will come out soft.
Light it evenly and watch for your own shadow, which is the commonest problem with hand-held document photos. Diffuse daylight near a window beats a ceiling light directly overhead. And lay the page flat - a curled page is not a flat surface, and no four-corner correction will flatten a curve.
After the correction
A corrected page often benefits from a contrast lift to get the paper closer to white and the text closer to black, which the Image Editor will do. If the document is going to be filed or emailed, converting it to a PDF with the Image to PDF tool makes it one file rather than a folder of photographs, and a multi-page document becomes a single document.
Common questions
Can it fix a crumpled page?
No. Four-corner correction assumes the page is flat. Creases and curl distort the text in ways the transformation cannot model, so flatten the paper first.
What if a corner is cut off in the photo?
Then there is no corner to mark, and the correction cannot be computed properly. Retake the photo with a margin around the page.
Will it make the text searchable?
No. This produces a clean image, not recognised text. Extracting text is a separate job - the PDF Text Extractor works on PDFs that already contain text rather than on pictures of text.
Is my document uploaded anywhere?
No. The correction runs in your browser, which matters more than usual when the document is a contract or an ID.
Can I scan several pages?
One at a time, then combine them. Correct each page, convert to PDF, and merge the results.
Got a photo of a page that needs straightening?