Seeing through Optical Scattering Media using Artificial Neural Networks

We recently discovered at the Langevin Institute that there exist subtle correlations between intensity patterns transmitted through and reflected off a scattering medium [Fayard 2019, PRL].

Such an effect allows us to obtain information about the transmitted photons by only measuring the reflected ones, opening new roads for non-invasive imaging procedures.

However, the imaging resolution scales as the thickness of the scattering medium, similar to diffuse tomography approaches, which is typically much larger than the wavelength.

Ideally, one wants to obtain information about the details of the transmitted field with a resolution of the order of the wavelength by only measuring reflected photons.

The novel approach consists in using Artificial Neural Networks to identify the relevant information hidden inside the reflected field about a target behind a scattering medium.

More information :

PDF - 530 ko

Haut de page