Persian Document Scanner with OCR and Automatic Orientation Correction

Description: This project leverages OpenCV and Tesseract OCR to detect and scan documents from images. It locates document contours, corrects perspective for a clean, top-down view, and applies OCR (using Persian language support) to recognize and extract text. The script also detects and adjusts image orientation based on OCR confidence, ensuring proper readability.

Requirements:

• Python 3.6+
• OpenCV for image processing (cv2 module)
• Pytesseract for Optical Character Recognition (OCR)

Python Libraries:

• numpy : For numerical operations
• cv2 : OpenCV library for image processing
• pytesseract : Tesseract OCR wrapper

Steps to Run:

1. Install Python libraries:

pip install numpy opencv-python pytesseract

2. Install Tesseract OCR:

• Windows: Download and install Tesseract from https://github.com/UB-Mannheim/tesseract/wiki.
• MacOS/Linux: Install Tesseract via package managers, e.g., brew install tesseract (Mac) or sudo apt install tesseract-ocr (Linux).

Usage

1. Prepare Input Image: Save the document image as input_image.jpg in the same directory (You can use PNG file).
2. Run the Python script app.py using the command:

python app.py

3. Output:

• result_image.jpg : Scanned document image after perspective correction.
• rotated_image.jpg : Final image with corrected orientation.

How It Works:

1. Edge Detection and Contour Detection:

• • Converts the image to grayscale, applies Gaussian blur, and performs Canny edge detection.
• • Identifies document boundaries and applies morphological operations to close gaps.

2. Perspective Correction:

• • Locates a contour with four corners (assumed to be the document) and applies perspective transformation to get a top-down view.

3. OCR and Orientation Detection:

• • Runs OCR on the center portion of the image at various angles (0°, 90°, 180°, and 270°).
• • Chooses the angle with the highest OCR confidence for final orientation.

Notes

• This script is optimized for Persian language OCR; modify the lang parameter in ocr_image if using other languages.
• Ensure pytesseract.pytesseract.tesseract_cmd points to your Tesseract installation if running on Windows.