LION.losses.SUREpgImage¶

Classes

SUREpgLoss(zeta, sigma2[, eps1, eps2, kappa])

PG-SURE: an unbiased estimate of a denoiser's mean squared error under mixed Poisson-Gaussian noise (Le Montagner, Angelini and Olivo-Marin, "An unbiased risk estimator for image denoising in the presence of mixed Poisson-Gaussian noise", IEEE TIP 23(3), 2014), as a self-supervised loss.

class LION.losses.SUREpgImage.SUREpgLoss(zeta, sigma2, eps1=0.001, eps2=0.001, kappa=1.0)[source]¶

Bases: object

PG-SURE: an unbiased estimate of a denoiser’s mean squared error under mixed Poisson-Gaussian noise (Le Montagner, Angelini and Olivo-Marin, “An unbiased risk estimator for image denoising in the presence of mixed Poisson-Gaussian noise”, IEEE TIP 23(3), 2014), as a self-supervised loss. zeta: the Poisson gain; sigma2: the Gaussian variance; eps1, eps2: the Monte Carlo perturbations of the first and second derivatives; kappa: the asymmetry of the second’s random perturbation.