Abstract

Speckle artifacts degrade image quality in virtually all modalities thatutilize coherent energy, including optical coherence tomography, reflectanceconfocal microscopy, and ultrasound. We present a deep learning framework forlaser speckle reduction, DeepLSR (https://durr.jhu.edu/DeepLSR), thattransforms coherently-illuminated images to a speckle-free domain. We applythis method to widefield images of objects and tissues illuminated with amulti-wavelength laser, reconstructing speckle-reduced images with greaterfidelity than conventional methods.

 

Quick Read (beta)