. Author manuscript; available in PMC: 2022 May 13.

Published in final edited form as: Proc IEEE Comput Soc Conf Comput Vis Pattern Recognit. 2021 Nov 13;2021:2423–2432. doi: 10.1109/cvpr46437.2021.00245

Table 2.

Quantitative comparison with models trained by different strategies in Scenario 2.

Methods	Data Centers (Institutions)		T₁-weighted				T₂-weighted
	Tain	Test	SSIM	PSRN	Average		SSIM	PSNR	Average
	Tain	Test	SSIM	PSRN	SSIM	PSNR	SSIM	PSNR	SSIM	PSNR
Single	B	B	0.9660	37.30	0.9351	33.81	0.9558	34.90	0.9278	32.35
	F	F	0.9494	35.45			0.9404	32.43
	H	H	0.8855	29.67			0.9001	31.29
	I	I	0.9396	32.80			0.9151	30.79
FL-MR	B, F, H, I	B	0.9662	37.37	0.9294	33.92	0.9482	35.34	0.9238	32.64
		F	0.9404	35.25			0.9306	32.19
		H	0.8732	30.03			0.9021	31.74
		I	0.9379	33.03			0.9145	31.29
FL-MRCM	B, F, H, I	B	0.9676	37.57	0.9381	34.14	0.9630	35.85	0.9373	33.13
		F	0.9475	35.57			0.9385	0.32.69
		H	0.8940	30.27			0.9232	32.44
		I	0.9432	33.13			0.9244	31.54
Mix (Upper Bound)	B, F, H, I	B	0.9698	37.62	0.9440	34.35	0.9655	5.83	0.9398	33.14
		F	0.9558	36.15			0.9435	32.82
		H	0.9047	30.57			0.9236	32.47
		I	0.9454	33.08			0.9266	31.44