notesum.ai

Published at December 10

Modeling Dual-Exposure Quad-Bayer Patterns for Joint Denoising and Deblurring

eess.IV

cs.CV

Released Date: December 10, 2024

Authors: Yuzhi Zhao¹, Lai-Man Po¹, Xin Ye², Yongzhe Xu², Qiong Yan²

Aff.: ¹City University of Hong Kong; ²SenseTime Research

Arxiv: http://arxiv.org/pdf/2412.07256v1

Refer to caption

Method	Input data structure	MACs	Running time	$K=0.25$ , $\sigma=5$		$K=0.5$ , $\sigma=5$		$K=0.75$ , $\sigma=5$
Method	Input data structure	(512 $\times$ 512)	(sec / 512 $\times$ 512)	PSNR	SSIM	PSNR	SSIM	PSNR	SSIM
LSD2 [11]	dual-exposure Quad-Bayer	113.433G	0.0079	32.21	0.9264	32.67	0.9317	32.81	0.9333
MWCNN [58]	dual-exposure Quad-Bayer	301.671G	0.1514	31.85	0.9193	32.28	0.9244	32.47	0.9278
SGN [91]	dual-exposure Quad-Bayer	51.258G	0.0113	32.00	0.9227	32.46	0.9281	32.64	0.9299
MIRNet [62]	dual-exposure Quad-Bayer	2.607T	0.6869	32.18	0.9275	32.72	0.9326	32.87	0.9339
MPRNet [31]	dual-exposure Quad-Bayer	2.642T	0.1435	32.24	0.9267	32.41	0.9301	32.81	0.9323
SRN [29]	dual-exposure Quad-Bayer	343.004G	0.0545	32.42	0.9296	32.86	0.9345	33.01	0.9358
DeblurGAN [8]	dual-exposure Quad-Bayer	205.445G	0.0111	32.55	0.9303	32.93	0.9336	33.15	0.9360
DMPHN [30]	dual-exposure Quad-Bayer	202.458G	0.0995	32.51	0.9303	32.93	0.9344	33.10	0.9359
MIMOUNet [10]	dual-exposure Quad-Bayer	300.205G	0.0394	32.71	0.9332	33.09	0.9366	33.31	0.9387
MIMOUNet++ [10]	dual-exposure Quad-Bayer	687.105G	0.0831	32.82	0.9350	33.26	0.9386	33.40	0.9398
QRNet	dual-exposure Quad-Bayer	34.628G	0.0178	33.24	0.9403	33.61	0.9428	33.81	0.9445