Document Scanner App

A simple computer vision project that scans documents from images and converts them into clean, readable PDF files.

Built using OpenCV and Streamlit.

---

Features

• Detects document edges automatically
• Corrects perspective (flattens the page)
• Removes shadows and background noise
• Enhances readability (black & white scan)
• Supports multiple images
• Exports scanned pages as a single PDF

---

Tech Stack

• Python
• OpenCV → image processing
• NumPy → numerical operations
• Streamlit → web app interface
• Pillow → PDF generation

---

How It Works

The app follows this pipeline:

1. Image Upload

   • User uploads one or more images.

2. Document Detection

   • Uses contour detection to find the document edges.
   • Filters shapes to detect a 4-sided boundary.

3. Perspective Transform

   • Converts tilted document into a flat top-down view.

4. Preprocessing

   • Converts to grayscale
   • Removes shadows using background normalization
   • Applies denoising

5. Thresholding

   • Converts image into clean black & white for readability

6. PDF Generation

   • All processed images are combined into a single PDF

---

Project Structure

doc_scanner/
├── app.py
├── src/
│   ├── scanner.py
│   ├── utils.py
├── data/
│   ├── input/
│   ├── output/
├── requirements.txt
├── README.md

---

Installation

1. Clone the repository

git clone https://github.com/YOUR_USERNAME/doc_scanner.git
cd doc_scanner

---

2. Create environment (conda recommended)

conda create --prefix ./venv python=3.10
conda activate ./venv

---

3. Install dependencies

pip install -r requirements.txt

---

Run the App

streamlit run app.py

Then open the link shown in terminal.

---

Usage

1. Upload one or more images

2. The app will automatically:

   • detect the document
   • scan and enhance it

3. Click Download PDF to get the final output

---

Key Concepts Used

1. Edge Detection

Detects boundaries of objects in the image.

2. Contour Detection

Finds shapes and identifies the document region.

3. Perspective Transform

Maps the document into a flat rectangular view.

4. Image Enhancement

• Noise removal
• Shadow removal
• Contrast improvement

5. Thresholding

Converts image into high-contrast black & white.

---

Future Improvements

• Live camera scanning
• Automatic cropping
• OCR (text extraction)

---

Author

Ahan Mondal