XPAR-Net enhances lens-free on-chip microscopy with cross-paradigm learning

XPAR-Net is a new network designed to improve lens-free on-chip microscopy by addressing the challenges of phase inference and fringe-scale information recovery from a single in-line hologram. It uses coordinated learning across different representation paradigms, combining field-wide channel-relation modeling and localized spatial refinement to enhance image reconstruction fidelity. This approach also includes cross-domain alignment and fusion to optimize spectral residual correction within the acquisition support.

Originally published by gnews