notesum.ai

Published at December 6

Learning to Translate Noise for Robust Image Denoising

eess.IV

cs.CV

Released Date: December 6, 2024

Authors: Inju Ha¹, Donghun Ryou¹, Seonguk Seo¹, Bohyung Han¹

Aff.: ¹Computer Vision Laboratory, Seoul National University

Arxiv: http://arxiv.org/pdf/2412.04727v1

Denoising Method	Metric	SIDD	Poly	CC	HighISO	iPhone	Huawei	OPPO	Sony	Xiaomi	OOD Avg.
		In-distribution	Out-of-distribution
R2R* [27]	PSNR	35.06	36.81	35.26	37.33	39.19	38.29	39.32	41.46	35.36	37.88
R2R* [27]	SSIM	0.9150	0.9722	0.9756	0.9712	0.9606	0.9663	0.9739	0.9729	0.9664	0.9699
AP-BSN* [15]	PSNR	36.32	35.88	33.13	36.66	39.82	37.01	39.04	40.04	33.37	36.87
AP-BSN* [15]	SSIM	0.9281	0.9751	0.9732	0.9777	0.9766	0.9628	0.9746	0.9798	0.9548	0.9718
SSID* [17]	PSNR	37.39	37.13	34.93	38.24	40.85	37.22	39.10	42.89	34.25	38.08
SSID* [17]	SSIM	0.9338	0.9799	0.9805	0.9808	0.9816	0.9652	0.9738	0.9878	0.9562	0.9757
Restormer* [36]	PSNR	39.93	37.63	36.31	38.24	40.05	38.36	39.49	44.02	35.62	38.85
Restormer* [36]	SSIM	0.9598	0.9790	0.9805	0.9753	0.9727	0.9671	0.9768	0.9889	0.9706	0.9745
Restormer-ours	PSNR	39.08	38.74	37.60	40.06	41.62	39.68	40.55	44.12	36.14	39.81
Restormer-ours	SSIM	0.9558	0.9846	0.9861	0.9851	0.9751	0.9761	0.9794	0.9849	0.9747	0.9807
NAFNet* [6]	PSNR	39.88	37.13	35.81	38.27	40.16	37.68	39.57	43.48	34.97	38.38
NAFNet* [6]	SSIM	0.9594	0.9714	0.9809	0.9784	0.9700	0.9676	0.9781	0.9824	0.9681	0.9746
NAFNet-ours	PSNR	39.17	38.67	37.82	39.94	41.94	39.74	40.45	44.17	36.14	39.86
NAFNet-ours	SSIM	0.9566	0.9851	0.9876	0.9853	0.9805	0.9778	0.9796	0.9869	0.9745	0.9822
KBNet* [42]	PSNR	40.26	36.79	35.21	38.05	37.93	35.14	37.73	41.65	34.23	37.09
KBNet* [42]	SSIM	0.9618	0.9785	0.9808	0.9785	0.9526	0.9459	0.9657	0.9784	0.9640	0.9681
KBNet-ours	PSNR	39.06	38.57	37.59	39.83	41.63	39.71	40.46	44.04	36.04	39.73
KBNet-ours	SSIM	0.9559	0.9840	0.9859	0.9845	0.9752	0.9773	0.9794	0.9839	0.9739	0.9805
XFormer* [38]	PSNR	39.89	37.41	35.93	37.82	39.92	38.30	39.54	43.74	35.52	38.52
XFormer* [38]	SSIM	0.9599	0.9774	0.9781	0.9738	0.9725	0.9673	0.9773	0.9886	0.9702	0.9756
XFormer-ours	PSNR	38.98	38.67	37.73	40.22	42.04	39.79	40.61	44.26	36.23	39.94
XFormer-ours	SSIM	0.9549	0.9852	0.9857	0.9857	0.9816	0.9773	0.9805	0.9879	0.9748	0.9824