Table 3.

SSIM metric values for UN-4 layer, Dense-UN and R2-UN networks concerning different noise distribution types with certain parameters – Gaussian, S&P, Speckle and Poisson.

Noise type	Noise Details	UN-4 layer	Dense-UN	R2-UN
Gaussian	${\sigma }^2=0.01$	$0.76\pm$0.035	0.83$\pm \mathit{0}.\mathit{019}$	0.80$\pm$0.015
	${\sigma }^2=0.02$	$0.75\pm$0.036	$\mathit{0}.\mathit{82}\pm \mathit{0}.\mathit{034}$	$0.78\pm 0.039$
	${\sigma }^2=0.04$	$0.72\pm 0.026$	$\mathit{0}.\mathit{80}\pm \mathit{0}.\mathit{028}$	$0.76\pm 0.035$
	${\sigma }^2=0.08$	$0.69\pm 0.028$	$\mathit{0}.\mathit{79}\pm \mathit{0}.\mathit{029}$	$0.74\pm 0.033$
S&P	Pixel destruction = 5 %	$0.60\pm 0.020$	$\mathit{0}.\mathit{81}\pm \mathit{0}.\mathit{018}$	$0.80\pm 0.021$
	Pixel destruction = 10 %	$0.51\pm 0.039$	$\mathit{0}.\mathit{80}\pm \mathit{0}.\mathit{03}$	$0.77\pm 0.037$
Speckle	${\sigma }^2=0.05$	$0.77\pm 0.033$	$\mathit{0}.\mathit{84}\pm \mathit{0}.\mathit{03}$	$0.80\pm 0.031$
	${\sigma }^2=0.1$	$0.74\pm 0.033$	$\mathit{0}.\mathit{83}\pm \mathit{0}.\mathit{027}$	$0.78\pm 0.036$
Poisson	$P\left(N\right)=\exp (-⟨N⟩)\frac{{⟨N⟩}^N}{N!}$	$0.76\pm 0.015$	$\mathit{0}.\mathit{83}\pm \mathit{0}.\mathit{023}$	$0.78\pm 0.035$