notesum.ai

Published at November 18

Zoomed In, Diffused Out: Towards Local Degradation-Aware Multi-Diffusion for Extreme Image Super-Resolution

cs.CV

cs.AI

cs.MM

Released Date: November 18, 2024

Authors: Brian B. Moser¹, Stanislav Frolov¹, Tobias C. Nauen¹, Federico Raue¹, Andreas Dengel¹

Aff.: ¹German Research Center for Artificial Intelligence, University of Kaiserslautern-Landau

Arxiv: http://arxiv.org/abs/2411.12072v1

Refer to caption

Type	Methods	PSNR $\uparrow$	SSIM $\uparrow$	LPIPS $\downarrow$
Regression	Bicubic	26.70	0.77	0.409
	EDSR	28.98	0.83	0.270
	RRDB	29.44	0.84	0.253
	CAR	32.82	0.88	-
GANs	RankSRGAN	26.55	0.75	0.128
GANs	ESRGAN	26.22	0.75	0.124
NFs	SRFlow	27.09	0.76	0.120
Diffusion	SR3 + YODA	27.24	0.77	0.127
(SR-trained)	SRDiff	27.41	0.79	0.136
	DiWa	28.09	0.78	0.104
Diffusion	SeeSR + MD	24.28	0.67	0.110
(T2I-trained)	Ours	24.34	0.68	0.108