Table 3 Quantitative comparisons between...

Table 3

Quantitative comparisons between MSAN and other methods in terms of the mean PSNR and SSIM on the FMD dataset [54]. The first and second best performances are denoted in red and blue, respectively. Methods marked with ‘*’ were trained with the training dataset [54].

Methods		PSNR(dB)/SSIM
		The number of raw images for averaging					Average value
		1	2	4	8	16	Average value
Traditional method	Noised	27.22 / 0.5442	30.08 / 0.6800	32.86 / 0.7981	36.03 / 0.8892	39.70 / 0.9487	33.18 / 0.7740
	NLM	31.25 / 0.7503	32.85 / 0.8116	34.92 / 0.8763	37.09 / 0.9208	40.04 / 0.9540	35.23 / 0.8626
	BM3D	32.71 / 0.7922	34.09 / 0.8430	36.05 / 0.8970	38.01 / 0.9336	40.61 / 0.9598	36.29 / 0.8851
	KSVD	32.02 / 0.7746	33.69 / 0.8327	35.84 / 0.8933	37.79 / 0.9314	40.36 / 0.9585	35.94 / 0.8781
	EPLL	32.61 / 0.7876	34.07 / 0.8414	36.08 / 0.8970	38.12 / 0.9349	40.83 / 0.9618	36.34 / 0.8845
	WNNM	32.52 / 0.7880	34.04 / 0.8419	36.04 / 0.8973	37.95 / 0.9334	40.45 / 0.9587	36.20 / 0.8839
	PURE-LET	31.95 / 0.7664	33.49 / 0.8270	35.29 / 0.8814	37.25 / 0.9212	39.59 / 0.9450	35.51 / 0.8682
Early DL	DnCNN*	34.88 / 0.9063	36.02 / 0.9257	37.57 / 0.9460	39.28 / 0.9588	41.57 / 0.9721	37.86 / 0.9418
	IRCNN*	34.70 / 0.8977	35.83 / 0.9217	37.37 / 0.9439	39.10 / 0.9571	41.18 / 0.9695	37.64 / 0.9380
	MemNet*	33.04 / 0.8314	35.23 / 0.9018	37.16 / 0.9383	39.02 / 0.9555	41.15 / 0.9687	37.12 / 0.9191
SotA DL	Noise2Noise*	35.40 / 0.9187	36.40 / 0.9230	37.59 / 0.9481	39.43 / 0.9601	41.45 / 0.9724	38.05 / 0.9445
	MWCNN*	35.40 / 0.9190	36.33 / 0.9329	37.62 / 0.9489	39.32 / 0.9608	41.39 / 0.9736	38.01 / 0.9470
	RIDNet*	35.63 / 0.9167	36.41 / 0.9325	37.97 / 0.9498	39.55 / 0.9610	41.58 / 0.9740	38.23 / 0.9468
	DPDN*	35.64 / 0.9189	36.35 / 0.9322	38.02 / 0.9501	39.51 / 0.9611	41.50 / 0.9744	38.20 / 0.9472
	WF-UNet*	34.45 / 0.8978	35.58 / 0.9204	37.29 / 0.9427	38.97 / 0.9561	41.23 / 0.9689	37.50 / 0.9372
	MSAN* (Ours)	35.78 / 0.9216	36.85 / 0.9362	38.19 / 0.9507	39.70 / 0.9621	41.31 / 0.9738	38.37 / 0.9489

Methods		PSNR(dB)/SSIM
		The number of raw images for averaging					Average value
		1	2	4	8	16	Average value
Traditional method	Noised	27.22 / 0.5442	30.08 / 0.6800	32.86 / 0.7981	36.03 / 0.8892	39.70 / 0.9487	33.18 / 0.7740
	NLM	31.25 / 0.7503	32.85 / 0.8116	34.92 / 0.8763	37.09 / 0.9208	40.04 / 0.9540	35.23 / 0.8626
	BM3D	32.71 / 0.7922	34.09 / 0.8430	36.05 / 0.8970	38.01 / 0.9336	40.61 / 0.9598	36.29 / 0.8851
	KSVD	32.02 / 0.7746	33.69 / 0.8327	35.84 / 0.8933	37.79 / 0.9314	40.36 / 0.9585	35.94 / 0.8781
	EPLL	32.61 / 0.7876	34.07 / 0.8414	36.08 / 0.8970	38.12 / 0.9349	40.83 / 0.9618	36.34 / 0.8845
	WNNM	32.52 / 0.7880	34.04 / 0.8419	36.04 / 0.8973	37.95 / 0.9334	40.45 / 0.9587	36.20 / 0.8839
	PURE-LET	31.95 / 0.7664	33.49 / 0.8270	35.29 / 0.8814	37.25 / 0.9212	39.59 / 0.9450	35.51 / 0.8682
Early DL	DnCNN*	34.88 / 0.9063	36.02 / 0.9257	37.57 / 0.9460	39.28 / 0.9588	41.57 / 0.9721	37.86 / 0.9418
	IRCNN*	34.70 / 0.8977	35.83 / 0.9217	37.37 / 0.9439	39.10 / 0.9571	41.18 / 0.9695	37.64 / 0.9380
	MemNet*	33.04 / 0.8314	35.23 / 0.9018	37.16 / 0.9383	39.02 / 0.9555	41.15 / 0.9687	37.12 / 0.9191
SotA DL	Noise2Noise*	35.40 / 0.9187	36.40 / 0.9230	37.59 / 0.9481	39.43 / 0.9601	41.45 / 0.9724	38.05 / 0.9445
	MWCNN*	35.40 / 0.9190	36.33 / 0.9329	37.62 / 0.9489	39.32 / 0.9608	41.39 / 0.9736	38.01 / 0.9470
	RIDNet*	35.63 / 0.9167	36.41 / 0.9325	37.97 / 0.9498	39.55 / 0.9610	41.58 / 0.9740	38.23 / 0.9468
	DPDN*	35.64 / 0.9189	36.35 / 0.9322	38.02 / 0.9501	39.51 / 0.9611	41.50 / 0.9744	38.20 / 0.9472
	WF-UNet*	34.45 / 0.8978	35.58 / 0.9204	37.29 / 0.9427	38.97 / 0.9561	41.23 / 0.9689	37.50 / 0.9372
	MSAN* (Ours)	35.78 / 0.9216	36.85 / 0.9362	38.19 / 0.9507	39.70 / 0.9621	41.31 / 0.9738	38.37 / 0.9489

[ViewLarge]