This pipeline converts a tilted, perspective-distorted photo into a clean A4-aligned scanned document.
This project establishes a robust document scanning pipeline using Python and OpenCV. It is designed to process smartphone-captured, perspective-distorted document photos, automatically cleaning and rectifying them to produce a clean, printable output. The pipeline performs the following core actions: Noise Reduction: The pipeline applies Median Blur for noise reduction because the input images often contain salt-and-pepper noise—a common artifact in smartphone-captured photos. Edge Detection: Employs Canny and Dilation to find and solidify the document's boundaries. Smart Detection: Detects and filters contours by area, using approxPolyDP to mathematically identify the four main corner vertices of the largest shape (the paper). Correction Logic: Mathematically sorts the corners and calculates the true dimensions (width/height) via Euclidean distance. Transformation: Applies the Warp Perspective to flatten the document, explicitly enforcing the standard A4 aspect ratio (1:1.414) for a professional, print-ready scan.
Python 3.10+ is recommended to run this project.
# Create virtual environment (optional but recommended)
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
# Install required packages
pip install -r requirements.txtThe table below shows the full transformation pipeline: original → noisy edges → cleaned edges → final perspective-corrected output.
| Original Image | Canny Detection (with noise) | Canny + Dilation (cleaned) | Final Output |
|---|---|---|---|
 |
 |
 |
 |
This project is licensed under the MIT License - see the LICENSE file for details.
Run the project by simply executing:
python main.py