Transformer improves wireless capsule endoscopy image resolution efficiently
CEM-TUDASR: Computationally efficient multi-modality transformer based unsupervised domain adaptive super-resolution approach
Computer Vision and Pattern Recognition
Summary
Wireless Capsule Endoscopy takes pictures inside the gut but the images are often blurry and low quality because of small cameras and wireless limits. The authors created a special computer program called CEM-TUDASR that makes these images clearer without needing examples of exact blurry and sharp pictures paired together for training. It uses modern techniques called Transformers to pay attention to details and the bigger picture at once. Tests show it works better than older methods and can also improve other medical images like retinal scans, all while running efficiently on limited hardware.
What this means in practice
- •For clinical imaging teams: Enhance the quality of capsule endoscopy images without extra paired training data, enabling better detection of digestive system details.
- •For embedded device developers: Integrate a lightweight and efficient Transformer-based super-resolution model into resource-constrained endoscopic hardware for real-time image improvement during procedures.
Authors
Anjali Sarvaiya, Jay Kadel, Kishor Upla, Kiran Raja
Abstract
Wireless Capsule Endoscopy (WCE) enables non-invasive visualization of the gastrointestinal tract, but its miniaturized optics, sensor limitations, and wireless transmission constraints result in low-resolution images with reduced visibility of diagnostically important structures. This paper proposes CEM-TUDASR, a computationally efficient unsupervised Transformer-based super-resolution framework for WCE image enhancement without paired low-resolution (LR) and high-resolution (HR) training data. A domain-adaptive degradation network synthesizes realistic WCE-like LR images from HR conventional endoscopy images, reducing the domain gap and enabling effective unpaired learning. The SR generator integrates Deep Attention Blocks (DABs) and a Fusion Attention Block (FAB) to capture long-range contextual dependencies and fine local structures while preserving perceptual and structural fidelity. The model is trained on a curated dataset derived from Kvasir Capsule and evaluated on KID and GIANA for cross-dataset generalization. No-reference quality metrics, including BRISQUE, PIQE, NIQE, and the domain-specific EndoQM, show that CEM-TUDASR consistently outperforms existing unsupervised SR methods. Qualitative results further demonstrate improved restoration of mucosal textures, vascular patterns, and clinically relevant anatomical details. Cross-domain experiments on retinal images additionally demonstrate the adaptability of the framework. With only 2.67 million parameters and 169.94 GFLOPs, CEM-TUDASR achieves high-quality reconstruction while maintaining computational efficiency, making it suitable for resource-constrained clinical and embedded endoscopic applications.