Two-Stage CNN Model for Joint Demosaicing and Denoising of Burst Bayer Images

Table 1

Comparison of multiframe joint demosaicing and denoising results .

Dataset	Noise level	BDNet		Interp.+V-BM4D		M2M		TwoStage
		PSNR	SSIM	PSNR	SSIM	PSNR	SSIM	PSNR	SSIM
Kodak		20.29	0.513	33.05	0.9213	35.90	0.9429	36.33	0.9487
		19.47	0.4789	28.63	0.7548	31.45	0.8623	31.88	0.8774
		20.49	0.5107	25.76	0.6161	28.90	0.7783	29.77	0.8198
McM		16.57	0.4417	32.99	0.9010	35.24	0.9360	34.92	0.9284
		18.82	0.4995	28.98	0.7630	31.40	0.8713	31.53	0.8732
		17.37	0.4564	26.16	0.6436	29.01	0.7941	29.67	0.8297
BSD 500		19.02	0.4492	32.80	0.9310	34.98	0.9524	35.41	0.9544
		18.88	0.4389	28.36	0.7845	30.47	0.8689	30.92	0.8805
		19.12	0.4432	25.53	0.6585	27.95	0.7799	28.75	0.8183