Project idea · ungraded

Document Scanner

Turn a phone photo of a page lying on a desk into a flat, black-and-white scan: find the page, straighten it, and clean up the lighting — what scanner apps do.

You will use

Milestones

Build the smallest working version first. Ticks are saved in this browser only.

01Load a photo, resize it to about 800 pixels wide and convert it to greyscale.
02Blur, run Canny and find the largest contour.
03Approximate the contour with cv2.approxPolyDP until it has four corners; draw them on the photo.
04Order the corners and warp the page flat with cv2.getPerspectiveTransform and cv2.warpPerspective.
05Clean the result with adaptive or Otsu thresholding and save it as a PNG.

Stretch goals

Handle photos where the page is not the largest object.
Process a folder of photos into one multi-page PDF.
Add a fallback that lets the user click the four corners when detection fails.

Data

Photograph a few printed pages on a contrasting desk with your phone (different angles and lighting). No dataset download is needed.

Run it yourself

pip install opencv-python numpy
python scan.py photo.jpg scanned.png