Learning-Based Dark and Blurred Underwater Image Restoration

Table 3

The test results of objective IQA methods between our algorithm and S + D + underwater enhancement and restoration algorithms.

Image ID	Assessment methods	(A)	(B)	(C)	(D)	(E)	(F)	(G)	(H) ours
100600	PSNR	13.271	13.360	15.058	12.287	11.253	11.273	11.265	27.036
100600	SSIM	0.371	0.434	0.556	0.297	0.094	0.095	0.094	0.840
100600	IL-NIQE	59.899	50.375	47.866	49.569	77.211	75.689	74.244	39.235
100600	NIQE	6.442	6.952	5.757	6.376	6.458	6.500	6.470	5.756
100602	PSNR	13.825	13.974	15.241	12.233	11.900	11.702	11.416	24.579
100602	SSIM	0.426	0.520	0.571	0.338	0.243	0.210	0.171	0.842
100602	IL-NIQE	63.870	50.368	50.735	47.969	56.586	65.474	68.118	32.961
100602	NIQE	6.271	7.019	5.766	6.312	6.453	6.175	6.727	5.9664
100603	PSNR	13.088	13.221	12.941	12.010	11.733	12.104	11.132	23.613
100603	SSIM	0.432	0.521	0.516	0.356	0.288	0.373	0.211	0.867
100603	IL-NIQE	63.646	39.312	39.850	39.882	43.502	40.647	42.096	37.503
100603	NIQE	5.988	6.281	5.621	5.679	5.867	5.693	5.998	5.622
100605	PSNR	21.213	22.223	17.194	17.472	16.744	16.845	15.717	22.167
100605	SSIM	0.617	0.744	0.690	0.577	0.524	0.537	0.452	0.770
100605	IL-NIQE	85.565	67.611	66.661	67.624	68.777	65.319	70.876	51.919
100605	NIQE	7.207	6.815	6.130	5.906	6.278	6.314	6.433	5.4618