SmartEditor-Interactive-OCR-with-Confidence-Filtering

SmartEditor is a Python-based OCR tool that combines traditional OCR with real-time user correction. It uses Tesseract OCR, confidence score filtering, and a Tkinter GUI to extract and edit text from scanned documents and noisy image files—reducing manual post-processing.

---

🧠 Key Features

• 📷 Image preprocessing: grayscale + Gaussian blur
• 🔍 Confidence-based word filtering (Tesseract)
• 🖼 Real-time OCR output with GUI editing (Tkinter)
• 💾 Saves user-edited text in .txt format
• 🎯 Tested on scanned documents, signage, forms, and handwritten images

---

🚀 How It Works

1. Upload a scanned image or photo-based document.
2. Preprocessing enhances text visibility (grayscale, denoise, threshold).
3. OCR & Filtering:
   • Tesseract extracts words and confidence scores.
   • Only words with confidence ≥ 60 are passed to the GUI.
4. GUI Interaction:
   • Preview annotated image with bounding boxes.
   • Edit extracted text live.
5. Export final corrected text to edited_text.txt.

---

🛠 Requirements

• Python 3.10+
• Tesseract OCR (Install & update path in script)
• Libraries:
  • pytesseract
  • opencv
  • python
  • numpy
  • tkinter
  • Pillow

⚙️ Installation & Usage

1. Install Tesseract Download: https://github.com/tesseract-ocr/tesseract Set path in script:

   pytesseract.pytesseract.tesseract_cmd = r"C:\Program Files\Tesseract-OCR\tesseract.exe"

2. Install Python dependencies

   pip install opencv-python numpy pytesseract Pillow

3. Run the application

   python SmartEditor-Interactive-OCR-with-Confidence-Filtering.py

📊 Experimental Results

Model	Word Accuracy	Character Accuracy
Default Tesseract	64.8%	77.8%
Convolutional Preprocess	—	61.6%
Super-resolution + Tess.	86.0%	89.7%
EasyOCR + Post-process	85–93%	—
SmartEditor (Proposed)	75%	73.0%

SmartEditor balances accuracy with user control, significantly reducing the burden of manual proofreading.