📄 Document Scanner with OpenCV

A Python-based document scanner that automatically detects document boundaries and creates a top-down perspective view of documents. This project is an implementation based on Adrian Rosebrock's tutorial from PyImageSearch, created as a learning exercise to understand computer vision concepts using OpenCV.

🎯 Project Purpose

This project serves as a hands-on learning experience for:

• Computer Vision fundamentals with OpenCV
• Image processing techniques (edge detection, contour finding, perspective transforms)
• Document processing automation
• Python image manipulation

✨ Features

• Automatic document detection using edge detection and contour analysis
• Perspective correction to get a top-down view of documents
• Image enhancement with adaptive thresholding for better readability
• Real-time visualization of each processing step
• Support for various image formats (JPG, PNG, etc.)

🛠️ Installation

Prerequisites

• Python 3.6 or higher
• pip (Python package installer)

Dependencies

Install the required packages:

pip install opencv-python
pip install scikit-image
pip install imutils
pip install numpy

Or install all dependencies at once:

pip install -r requirements.txt

📁 Project Structure

scan_py/
├── scan.py          # Main document scanning script
├── transform.py     # Perspective transform utilities
├── images/          # Sample images directory
│   └── receipt.jpg  # Example document image
└── README.md        # This file

🚀 Usage

Basic Usage

Run the scanner with an image file:

python scan.py -i path/to/your/image.jpg

Examples

1. Scan a receipt in the images folder:

   python scan.py -i images/receipt.jpg

2. Scan any image from your computer:

   python scan.py -i "C:/Users/YourName/Desktop/document.png"

3. Scan an image in the current directory:

   python scan.py -i my_document.jpg

🔍 How It Works

The document scanner follows these steps:

Step 1: Edge Detection

• Converts the image to grayscale
• Applies Gaussian blur to reduce noise
• Uses Canny edge detection to find document boundaries

Step 2: Contour Detection

• Finds all contours in the edge-detected image
• Sorts contours by area (largest first)
• Identifies the contour with exactly 4 points (document corners)

Step 3: Perspective Transform

• Applies a four-point perspective transform
• Creates a top-down view of the document
• Enhances the image with adaptive thresholding

🎓 Learning Objectives

This project demonstrates key computer vision concepts:

• Image Preprocessing: Grayscale conversion, blurring, edge detection
• Contour Analysis: Finding and filtering contours based on properties
• Geometric Transformations: Perspective correction using homography
• Image Enhancement: Adaptive thresholding for better contrast
• OpenCV Integration: Working with cv2 library for image processing

⚠️ Troubleshooting

Common Issues

1. "screenCnt is not defined" error:

   • The script couldn't find a document with 4 clear corners
   • Try with a clearer image or better lighting
   • Ensure the document is fully visible in the frame

2. No contours detected:

   • Check if the image has sufficient contrast
   • Try adjusting the Canny edge detection parameters
   • Ensure the document edges are clearly defined

3. Poor quality results:

   • Use images with good lighting
   • Ensure the document is flat and not wrinkled
   • Try different angles if the document isn't detected

🔧 Customization

You can modify the script to adjust:

• Edge detection sensitivity (Canny thresholds)
• Blur intensity (Gaussian blur kernel size)
• Contour approximation accuracy
• Thresholding parameters for final enhancement

📚 Resources for Learning

• Original Tutorial by Adrian Rosebrock - The source tutorial this implementation is based on
• OpenCV Documentation
• Computer Vision Tutorials
• Image Processing Concepts

🤝 Contributing

This is a learning project, but suggestions and improvements are welcome! Feel free to:

• Report issues
• Suggest improvements
• Share your own implementations

📄 License

This project is open source and available under the MIT License.

---

Happy Learning! 🎉

This project is an implementation based on Adrian Rosebrock's tutorial from PyImageSearch, created to explore computer vision concepts and document processing techniques using OpenCV.

🙏 Acknowledgments

• Adrian Rosebrock - Original tutorial author and creator of PyImageSearch
• PyImageSearch - For providing excellent computer vision tutorials and resources