Scan a Document with Your Camera or a Photo

The core of a document scanner is a homography: the 3×3 transform that maps the four corners you marked to a true rectangle. Once solved, every output pixel is traced back through the transform to its source position, which un-does the keystone distortion of a phone held at an angle. Output dimensions come from your marked quad's real edge lengths, so an A4 page yields A4 proportions without you declaring the paper size.

Corner placement is the entire skill, and it takes ten seconds: zoom your eyes to each corner and put the handle where the paper's edges would intersect, even when a finger or shadow covers the actual corner. Precision here is worth more than camera quality — a 12-megapixel photo with sloppy corners produces a worse scan than an 8-megapixel one with exact ones.

Document mode then handles the illumination problem. A photographed page is never evenly lit, so a single global threshold either drowns one side in black or fades the other to nothing. The cleanup here normalises each pixel against a heavily blurred copy of the page — approximating the local background level — before thresholding, which is why a page with a soft shadow across it still comes out uniformly white with solid text. The ink cutoff slider sets how aggressive that separation is.

Keep photos of ID cards, bank letters, contracts and medical papers in colour mode when stamps, holograms or signatures in blue matter — binarisation flattens colour evidence. And chain the result into the Image-to-PDF tool when a portal wants a PDF: photograph, straighten, clean, paginate, all without any of it touching a server.