Video Denoising Based on a Spatiotemporal Kalman-Bilateral Mixture Model

Table 1

PSNR and SSIM comparisons of video denoising algorithms for 4 video sequences at 5 noise levels.

Video sequence; noise std (𝜎)	Salesman					Hall
	10	15	20	50	100	10	15	20	50	100
PSNR results (dB)
ST-GSM [15]	37.93	35.56	33.89	26.43	20.72	38.28	35.99	34.12	27.16	19.99
VBM3D [13]	39.11	36.65	34.72	27.93	22.18	39.96	37.93	36.31	28.14	21.97
ST-KBM	35.48	33.81	33.52	29.64	22.73	35.73	33.21	32.84	28.44	23.42
SSIM results
ST-GSM [15]	0.970	0.950	0.928	0.699	0.452	0.975	0.965	0.955	0.882	0.620
VBM3D [13]	0.976	0.958	0.932	0.742	0.489	0.980	0.973	0.966	0.887	0.601
ST-KBM	0.954	0.942	0.934	0.864	0.734	0.978	0.970	0.967	0.929	0.830
Video Sequence; Noise std (𝜎)	Akiyo					Silent
	10	15	20	50	100	10	15	20	50	100
PSNR results (dB)
ST-GSM [15]	40.67	38.34	36.53	28.44	21.89	37.41	35.17	33.61	27.63	21.87
VBM3D [13]	42.00	39.72	37.85	30.69	23.36	38.67	36.34	34.59	28.38	23.08
ST-KBM	35.09	34.66	34.06	30.17	23.67	34.26	32.78	32.29	27.98	23.10
SSIM results
ST-GSM [15]	0.980	0.969	0.958	0.852	0.673	0.963	0.943	0.922	0.787	0.561
VBM3D [13]	0.984	0.976	0.964	0.871	0.614	0.970	0.951	0.928	0.773	0.535
ST-KBM	0.962	0.954	0.943	0.894	0.795	0.937	0.922	0.907	0.823	0.692