🖼️ Enhancing Image Restoration with Quantum‑Integrated Contrastive Adversarial Networks

---

📌 Overview

This repository contains the implementation of the research paper
"Enhancing Image Restoration with Quantum‑Integrated Contrastive Adversarial Networks",
presented at IEEE INDISCON‑2025.

The proposed model synergistically combines:

• Contrastive Adversarial Networks (CANs) for robust adversarial learning
• Quantum Neural Networks (QNNs) via CUDA‑Q for quantum‑enhanced feature extraction

This hybrid quantum‑classical framework addresses the limitations of traditional image restoration by:

• Capturing intricate image details through quantum‑enhanced convolutions
• Leveraging contrastive loss to improve restoration quality
• Accelerating computation with CUDA‑Q integration

---

🧪 Abstract

Our model was evaluated on 50 random CIFAR‑10 images, achieving:

• SSIM: 0.8873 (avg)
• PSNR: 22.5981 (avg)
• p‑value: < 0.0001 (t‑test vs. noisy baselines)

These results demonstrate statistically significant improvements in both perceptual and quantitative metrics, while reducing processing times and enhancing fine‑detail reconstruction.

---

✨ Key Features

• Hybrid Quantum‑Classical Architecture: PyTorch + CUDA‑Q
• Generator: U‑Net style with residual connections + quantum‑enhanced convolutions
• Discriminator: PatchGAN‑inspired with contrastive loss
• Training Data: Waterloo Exploration Database
• Testing Data: CIFAR‑10
• Metrics: SSIM, PSNR, statistical significance testing

---

📊 Results Summary

Metric	Value (avg)	Dataset	Notes
SSIM	0.8873	CIFAR‑10	High structural similarity
PSNR	22.5981	CIFAR‑10	Strong noise suppression
p‑value	< 0.0001	—	Statistically significant vs. baseline

Example outputs are available in the figures/ directory.
You can also run inference to generate new restored images.

---

⚙️ Installation

Prerequisites:

• Python 3.9+
• PyTorch
• CUDA‑Q

---

Visual Results

📬 Contact

Satvik Raghav 📧 satvikraghav007@gmail.com

🙏 Acknowledgments

• IEEE INDISCON‑2025 presentation

• Waterloo Exploration Database

• CIFAR‑10 Dataset

• CUDA‑Q for quantum computing integration

💡 Skills Demonstrated

• Quantum‑integrated deep learning

• Contrastive adversarial training

• Image restoration & enhancement

• Statistical performance evaluation

• Hybrid architecture design (classical + quantum)

• CUDA‑accelerated computation

Required Libraries:

pip install numpy pandas scikit-learn matplotlib