🚀 AI Image Enhancement Tool (Real-ESRGAN + OpenCV)

A modern desktop application for image enhancement and super-resolution, combining classical computer vision techniques with GAN-based deep learning upscaling.

Built with a clean UI using CustomTkinter, this tool provides both manual controls and deep learning enhancement in one workflow.

✨ Features

🎛️ Manual Image Enhancement

• Brightness adjustment (real-time slider)
• Contrast control
• Sharpening filter (custom kernel)
• Noise reduction (OpenCV denoising)

🤖 AI Super-Resolution

• GAN-based super-resolution (Real-ESRGAN, x4 upscaling)
• Runs on CPU (no GPU required)

🖥️ Modern UI

• Built with CustomTkinter
• Responsive layout (controls + preview panels)
• Real-time preview updates

💾 Export Options

• Save manual enhancement
• Save AI-enhanced image (PNG for best quality)
• Save both versions at once

🧠 Architecture Overview

Image-Enhancer-Tool/

app/ ├── core/ │ └── ops.py # Image processing logic (OpenCV + AI) │ ├── ui/ │ └── gui.py # UI (CustomTkinter) │ └── utils/ └── (model goes here)

main.py # Entry point requirements.txt README.md

⚙️ Installation

1. Clone Repository git clone https://github.com/N1N0u/Image-Enhancer-Tool.git cd Image-Enhancer-Tool

2. Install Dependencies pip install -r requirements.txt

📥 Model Download (Required)

This project uses a pretrained model from Real-ESRGAN.

The model file is NOT included in this repository due to:

• Licensing considerations
• Best practices for clean repositories

👉 Download it from the official source: https://github.com/xinntao/Real-ESRGAN/releases

After downloading, place it here: app/utils/realesr-general-x4v3.pth

▶️ Usage

Run the application:

python main.py

Workflow:

1. Load an image
2. Adjust brightness/contrast
3. Click "AI Enhance"
4. Save results

🧪 Technologies Used

• Python
• OpenCV
• PyTorch
• Real-ESRGAN
• CustomTkinter
• NumPy
• Pillow (PIL)

💻 Development Environment

This project was developed and tested on low-end hardware:

• CPU: Intel Core 2 Quad Q8300 @ 2.5GHz
• RAM: 4GB
• GPU: None (CPU-only, no VRAM)

⚙️ Optimization Approach

• Efficient OpenCV pipelines for real-time processing
• Lightweight AI model (SRVGGNetCompact)
• CPU-compatible PyTorch inference
• Memory-conscious image handling

Despite limitations, the system supports:

• Real-time manual enhancement
• AI-based super-resolution on CPU

Note: AI enhancement is slower on CPU and may take several seconds per image.

🚀 Performance Insight

• Manual enhancements: near-instant
• AI enhancement (CPU): ~5–20 seconds depending on image size

Future optimization ideas:

• Model quantization (INT8)
• ONNX optimization
• Background threading (avoid UI freeze)

🚧 Future Improvements

• Add multithreading (non-blocking UI)
• Batch image processing
• Drag & drop support
• Multiple AI models (x2, x4, anime, etc.)
• Export quality settings
• Convert to standalone .exe

📌 Key Highlights (For Recruiters)

• Combines classical computer vision + deep learning
• Clean architecture (UI separated from processing)
• Runs on low-spec hardware (CPU-only)
• Real-world usable application
• Demonstrates PyTorch model loading and inference

📄 License

MIT License

🙏 Acknowledgements

• Real-ESRGAN by xinntao (used for super-resolution model)
• Open-source contributors in computer vision and deep learning

👨‍💻 Author

ATEF Aliat

Passionate about AI, Computer Vision, and Software Engineering. Focused on building real-world tools combining ML + UI.