Regularized supervised Bayesian approach for image deconvolution with regularization parameter estimation

EURASIP Journal on Advances in Signal Processing

Table 1 PSNR and SSIM values obtained by five methods for seven different images degraded by kernel 1

Image	Size	Measure	Blurred	Method
				RL	Wiener	TV	BTV	MAP- H¹
Lena	256×256	PSNR	23.15	30.73	31.01	32.75	33.17	33.95
		SSIM	0.6731	0.8428	0.8647	0.9147	0.9202	0.9381
Cameraman	512×512	PSNR	24.66	34.90	35.22	36.27	36.37	37.67
		SSIM	0.7824	0.9251	0.9249	0.9527	0.9546	0.9651
House	256×256	PSNR	25.09	31.97	32.00	35.59	35.94	36.13
		SSIM	0.7379	0.8502	0.8569	0.9263	0.9292	0.9347
Couple	512×512	PSNR	24.17	33.48	33.77	34.11	34.49	34.89
		SSIM	0.6127	0.9139	0.9158	0.9373	0.9397	0.9458
Pirate	512×512	PSNR	24.75	33.75	33.55	33.80	34.24	34.70
		SSIM	0.6268	0.9114	0.9140	0.9231	0.9283	0.9365
Boat	512×512	PSNR	23.98	33.57	33.68	33.60	33.95	34.52
		SSIM	0.6181	0.9044	0.9051	0.9151	0.9161	0.9276
Fingerprint	512×512	PSNR	17.54	26.84	26.98	28.56	28.66	29.27
		SSIM	0.3292	0.9177	0.9189	0.9495	0.9507	0.9564

For each test setting, six results are provided: Blurred, Richardson-Lucy algorithm, Wiener filter, TV method, BTV approach, and our proposed model. Bold format: the best score in each line