On the Use of Spatial and Spectral Redundancy to Speed-Up Brillouin Micro-Imaging
Résumé
We propose a smart scanning protocol to speed up the acquisition of Brillouin microscopy. It is based on a two-pass algorithm. In a first pass at low signal-to-noise-ratio (SNR),essential pixels located on the convex hull of the Fourier phasor space of the sample are identified, selected and rescanned at high
SNR. These essential spectra serve as a decomposition basis for the remaining spectra located inside the convex hull. A second acceleration is gained through the use of superpixels computed on a denoised version of the low SNR image. It allows to reduce the number of pixels on which Brillouin spectra are to be estimated.
This protocol recently introduced for Raman microscopy is tested here for the first time on Brillouin hyperspectral images (HSI). A compression by a factor between 3 and 253 with very limited distortion is demonstrated on simulated samples fitted on real samples of cell bones. Further opportunities of improvement are identified and discussed.
Domaines
Imagerie médicaleOrigine | Fichiers éditeurs autorisés sur une archive ouverte |
---|---|
licence |
Copyright (Tous droits réservés)
|