Deep-learning-aided spectroscopic detection and fusion imaging using a micro-tapered fiber and multimodal speckles
Abstract
Multimode fibers are promising for compact imaging and spectroscopy. However, current implementations are typically limited to a single function and often lack robustness against environmental disturbances. Unlike approaches that solely analyze the fiber end-face, we exploit the high information density of the leaky field from a fiber taper. A lensless system with a deep learning framework is developed, simultaneously capturing multimodal data from the taper leaky field and the end-face speckle. This approach achieves spectral reconstruction with a resolution of 0.05 nm and enables high-quality image recovery. By fusing both light fields, we significantly enhance image quality (MNIST SSIM up to 0.99), demonstrating a robust, all-fiber platform for integrated spectroscopic and imaging applications.
Fields of Study & Subjects
Random lasers and scattering media
Optical Imaging and Spectroscopy Techniques
Digital Holography and Microscopy
Publication details
- Journal / source
- Photonics Research
- Publisher
- Optica Publishing Group
- Volume
- 14
- Issue
- 5
- Pages
- 1737–1737
- Publication date
- 27 January 2026
- Language
- EN