notesum.ai

Published at November 27

Hierarchical Information Flow for Generalized Efficient Image Restoration

cs.CV

Released Date: November 27, 2024

Authors: Yawei Li¹, Bin Ren², Jingyun Liang¹, Rakesh Ranjan³, Mengyuan Liu⁴, Nicu Sebe⁵, Ming-Hsuan Yang⁶, Luca Benini¹

Aff.: ¹ETH Zürich; ²University of Pisa; ³Meta Reality Labs; ⁴Peking University; ⁵University of Trento; ⁶University of California, Merced

Arxiv: http://arxiv.org/abs/2411.18588v1

Method	Scale	Params	Set5		Set14		BSD100		Urban100		Manga109
Method	Scale	[M]	PSNR $\uparrow$	SSIM $\uparrow$	PSNR $\uparrow$	SSIM $\uparrow$	PSNR $\uparrow$	SSIM $\uparrow$	PSNR $\uparrow$	SSIM $\uparrow$	PSNR $\uparrow$	SSIM $\uparrow$
EDSR (Lim et al., 2017)	$2\times$	40.73	38.11	0.9602	33.92	0.9195	32.32	0.9013	32.93	0.9351	39.10	0.9773
SRFBN (Li et al., 2019b)	$2\times$	2.14	38.11	0.9609	33.82	0.9196	32.29	0.9010	32.62	0.9328	39.08	0.9779
RCAN (Zhang et al., 2018b)	$2\times$	15.44	38.27	0.9614	34.12	0.9216	32.41	0.9027	33.34	0.9384	39.44	0.9786
SAN (Dai et al., 2019)	$2\times$	15.71	38.31	0.9620	34.07	0.9213	32.42	0.9028	33.10	0.9370	39.32	0.9792
HAN (Niu et al., 2020)	$2\times$	63.61	38.27	0.9614	34.16	0.9217	32.41	0.9027	33.35	0.9385	39.46	0.9785
NLSA (Mei et al., 2021)	$2\times$	42.63	38.34	0.9618	34.08	0.9231	32.43	0.9027	33.42	0.9394	39.59	0.9789
IPT (Chen et al., 2021)	$2\times$	115.48	38.37	-	34.43	-	32.48	-	33.76	-	-	-
SwinIR (Liang et al., 2021)	$2\times$	11.75	38.42	0.9623	34.46	0.9250	32.53	0.9041	33.81	0.9427	39.92	0.9797
CAT-A (Chen et al., 2022b)	$2\times$	16.46	38.51	0.9626	34.78	0.9265	32.59	0.9047	34.26	0.9440	40.10	0.9805
ART (Zhang et al., 2022)	$2\times$	16.40	38.56	0.9629	34.59	0.9267	32.58	0.9048	34.3	0.9452	40.24	0.9808
EDT (Li et al., 2021)	$2\times$	11.48	38.63	0.9632	34.80	0.9273	32.62	0.9052	34.27	0.9456	40.37	0.9811
GRL-B (Li et al., 2023a)	$2\times$	20.05	38.67	0.9647	35.08	0.9303	32.67	0.9087	35.06	0.9505	40.67	0.9818
HAT (Chen et al., 2023)	$2\times$	20.62	38.73	0.9637	35.13	0.9282	32.69	0.9060	34.81	0.9489	40.71	0.9819
Hi-IR-B (Ours)	$2\times$	14.68	38.71	0.9657	35.16	0.9299	32.73	0.9087	34.94	0.9484	40.81	0.9830
HAT-L (Chen et al., 2023)	$2\times$	40.70	38.91	0.9646	35.29	0.9293	32.74	0.9066	35.09	0.9505	41.01	0.9831
Hi-IR-L (Ours)	$2\times$	39.07	38.87	0.9663	35.27	0.9311	32.77	0.9092	35.16	0.9505	41.22	0.9846
EDSR (Lim et al., 2017)	$3\times$	43.68	34.65	0.9280	30.52	0.8462	29.25	0.8093	28.80	0.8653	34.17	0.9476
SRFBN (Li et al., 2019b)	$3\times$	2.83	34.70	0.9292	30.51	0.8461	29.24	0.8084	28.73	0.8641	34.18	0.9481
RCAN (Zhang et al., 2018b)	$3\times$	15.63	34.74	0.9299	30.65	0.8482	29.32	0.8111	29.09	0.8702	34.44	0.9499
SAN (Dai et al., 2019)	$3\times$	15.90	34.75	0.9300	30.59	0.8476	29.33	0.8112	28.93	0.8671	34.30	0.9494
HAN (Niu et al., 2020)	$3\times$	64.35	34.75	0.9299	30.67	0.8483	29.32	0.8110	29.10	0.8705	34.48	0.9500
NLSA (Mei et al., 2021)	$3\times$	45.58	34.85	0.9306	30.70	0.8485	29.34	0.8117	29.25	0.8726	34.57	0.9508
IPT (Chen et al., 2021)	$3\times$	115.67	34.81	-	30.85	-	29.38	-	29.49	-	-	-
SwinIR (Liang et al., 2021)	$3\times$	11.94	34.97	0.9318	30.93	0.8534	29.46	0.8145	29.75	0.8826	35.12	0.9537
CAT-A (Chen et al., 2022b)	$3\times$	16.64	35.06	0.9326	31.04	0.8538	29.52	0.8160	30.12	0.8862	35.38	0.9546
ART (Zhang et al., 2022)	$3\times$	16.58	35.07	0.9325	31.02	0.8541	29.51	0.8159	30.1	0.8871	35.39	0.9548
EDT (Li et al., 2021)	$3\times$	11.66	35.13	0.9328	31.09	0.8553	29.53	0.8165	30.07	0.8863	35.47	0.9550
GRL-B (Li et al., 2023a)	$3\times$	20.24	35.12	0.9353	31.27	0.8611	29.56	0.8235	30.92	0.8990	35.76	0.9566
HAT (Chen et al., 2023)	$3\times$	20.81	35.16	0.9335	31.33	0.8576	29.59	0.8177	30.7	0.8949	35.84	0.9567
Hi-IR-B (Ours)	$3\times$	14.87	35.11	0.9372	31.37	0.8598	29.60	0.8240	30.79	0.8977	35.92	0.9583
HAT-L (Chen et al., 2023)	$3\times$	40.88	35.28	0.9345	31.47	0.8584	29.63	0.8191	30.92	0.8981	36.02	0.9576
Hi-IR-L (Ours)	$3\times$	39.26	35.20	0.9380	31.55	0.8616	29.67	0.8256	31.07	0.9020	36.12	0.9588
EDSR (Lim et al., 2017)	$4\times$	43.09	32.46	0.8968	28.80	0.7876	27.71	0.7420	26.64	0.8033	31.02	0.9148
SRFBN (Li et al., 2019b)	$4\times$	3.63	32.47	0.8983	28.81	0.7868	27.72	0.7409	26.60	0.8015	31.15	0.9160
RCAN (Zhang et al., 2018b)	$4\times$	15.59	32.63	0.9002	28.87	0.7889	27.77	0.7436	26.82	0.8087	31.22	0.9173
SAN (Dai et al., 2019)	$4\times$	15.86	32.64	0.9003	28.92	0.7888	27.78	0.7436	26.79	0.8068	31.18	0.9169
HAN (Niu et al., 2020)	$4\times$	64.20	32.64	0.9002	28.90	0.7890	27.80	0.7442	26.85	0.8094	31.42	0.9177
NLSA (Mei et al., 2021)	$4\times$	44.99	32.59	0.9000	28.87	0.7891	27.78	0.7444	26.96	0.8109	31.27	0.9184
IPT (Chen et al., 2021)	$4\times$	115.63	32.64	-	29.01	-	27.82	-	27.26	-	-	-
SwinIR (Liang et al., 2021)	$4\times$	11.90	32.92	0.9044	29.09	0.7950	27.92	0.7489	27.45	0.8254	32.03	0.9260
CAT-A (Chen et al., 2022b)	$4\times$	16.60	33.08	0.9052	29.18	0.7960	27.99	0.7510	27.89	0.8339	32.39	0.9285
ART (Zhang et al., 2022)	$4\times$	16.55	33.04	0.9051	29.16	0.7958	27.97	0.7510	27.77	0.8321	32.31	0.9283
EDT (Li et al., 2021)	$4\times$	11.63	33.06	0.9055	29.23	0.7971	27.99	0.7510	27.75	0.8317	32.39	0.9283
GRL-B (Li et al., 2023a)	$4\times$	20.20	33.10	0.9094	29.37	0.8058	28.01	0.7611	28.53	0.8504	32.77	0.9325
HAT (Chen et al., 2023)	$4\times$	20.77	33.18	0.9073	29.38	0.8001	28.05	0.7534	28.37	0.8447	32.87	0.9319
Hi-IR-B (Ours)	$4\times$	14.83	33.14	0.9095	29.40	0.8029	28.08	0.7611	28.44	0.8448	32.90	0.9323
HAT-L (Chen et al., 2023)	$4\times$	40.85	33.30	0.9083	29.47	0.8015	28.09	0.7551	28.60	0.8498	33.09	0.9335
Hi-IR-L (Ours)	$4\times$	39.22	33.22	0.9103	29.49	0.8041	28.13	0.7622	28.72	0.8514	33.13	0.9366