Ultra-low sampling fiber imaging system based on physical-enhanced IPSR
Abstract
Multimode fiber (MMF) imaging holds great potential in the field of minimally invasive endoscopy due to its ultra-fine outer diameter. However, the requirements for real-time monitoring and reduced phototoxicity force the system to operate at extremely low sampling rates, resulting in a severe loss of fidelity with traditional methods. Here, we introduce a physical-enhanced IPSR (Phys-IPSR) imaging framework. By implementing a deterministic front-end screening strategy to minimize mutual coherence and utilizing a dual-stage network to retrieve high-frequency features from extremely sparse data, Phys-IPSR significantly improves information encoding efficiency. At an extremely low sampling rate of only 0.77%, the system successfully achieves high-resolution reconstruction of 198 × 198 pixels, with an average PSNR of 36.46 dB and an SSIM of 0.8682, representing a fidelity improvement exceeding 20 dB over traditional compressed sensing methods. Furthermore, this framework demonstrates exceptional robustness within an axial defocus range of 30 mm (±15 mm), with a single-frame processing time of only 0.047 seconds. Phys-IPSR provides a design paradigm for high-fidelity fiber imaging under extreme sampling conditions, laying an important foundation for future clinically compatible endoscopic systems.
Fields of Study & Subjects
Random lasers and scattering media
Sparse and Compressive Sensing Techniques
Optical Imaging and Spectroscopy Techniques
Publication details
- Journal / source
- Optics Express
- Volume
- 34
- Issue
- 6
- Pages
- 10645–10645
- Publication date
- 17 February 2026
- Language
- EN