midorino555/flatbed-scanner-calibration

This project measures and corrects the geometric distortion of a consumer flatbed scanner, so that scanned paper drawings come out accurate enough to trace directly in CAD software. The author built it for scanning technical drawings into Fusion 360, but it works with any flatbed scanner, not just the specific model they used. Flatbed scanners are not perfectly accurate: straight lines can come out slightly bent, and the amount of distortion varies across the scan bed rather than being a simple, constant scaling error. That means the error cannot be fixed just by resizing or rotating the image afterward. This tool measures that distortion by scanning a printed grid of dots, fitting a mathematical correction model to how those dots are displaced from where they should be, and then applying that correction to future scans. In the author's testing, average error across an A3 page dropped from about 0.37 millimeters to about 0.04 millimeters after correction. Using it involves printing a calibration target at exactly 100 percent scale, carefully measuring the printed target itself with a ruler to account for any paper shrinkage from printing, scanning that target with all automatic corrections turned off, then running a Python script to build a calibration file specific to your scanner. Once that calibration file exists, a second script applies it to any new scan, saving both a corrected image and a backup of the original so nothing gets overwritten by accident. Windows users can drag a scan onto a batch file to apply the correction, and macOS users have an equivalent right-click Quick Action. Setup requires Python 3.9 or newer and a few Python packages installed via pip. The README stresses that a generated calibration file only works for the exact scanner it was made from and should not be shared or reused between different machines.