Learning-Based Dark and Blurred Underwater Image Restoration

Table 5

The test results of objective IQA methods between our algorithm and deep underwater image enhancement.

Image ID	Assessment methods	(A)	(B)	(C)	(D)	(E)	(F)	(G) ours
100600	PSNR	20.947	22.403	23.485	22.829	19.190	18.724	27.036
100600	SSIM	0.646	0.781	0.780	0.783	0.744	0.741	0.840
100600	IL-NIQE	58.020	48.487	54.521	59.034	81.410	72.060	39.235
100600	NIQE	7.334	6.732	6.903	7.652	10.062	12.704	5.756
100602	PSNR	21.935	16.091	24.409	21.828	9.366	18.593	24.579
100602	SSIM	0.757	0.655	0.813	0.749	0.470	0.736	0.842
100602	IL-NIQE	66.381	50.995	62.061	62.318	68.022	51.325	32.961
100602	NIQE	7.331	7.222	6.996	7.698	6.899	6.638	5.9664
100603	PSNR	17.440	20.641	18.465	17.364	12.476	16.236	23.613
100603	SSIM	0.643	0.807	0.736	0.651	0.615	0.761	0.867
100603	IL-NIQE	47.975	44.817	45.509	53.039	94.663	68.420	37.503
100603	NIQE	6.812	7.003	6.410	7.104	7.584	5.852	5.622
100605	PSNR	16.170	20.454	18.611	16.791	17.272	16.095	22.167
100605	SSIM	0.551	0.696	0.671	0.607	0.638	0.640	0.770
100605	IL-NIQE	71.732	44.950	58.549	73.383	100.597	95.610	51.919
100605	NIQE	6.848	6.373	6.163	7.693	8.905	8.598	5.4618