Learning-Based Dark and Blurred Underwater Image Restoration

Table 4

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

Image ID	Assessment methods	(A)	(B)	(C)	(D)	(E)	(F)	(H) ours
100600	PSNR	17.558	21.831	19.104	22.417	20.759	20.914	27.036
100600	SSIM	0.532	0.697	0.445	0.717	0.578	0.592	0.840
100600	IL-NIQE	68.850	61.838	68.343	66.967	61.455	60.936	39.235
100600	NIQE	7.200	6.909	6.814	7.166	6.131	6.135	5.756
100602	PSNR	12.693	20.232	18.472	20.502	18.872	18.844	24.579
100602	SSIM	0.423	0.671	0.541	0.622	0.551	0.544	0.842
100602	IL-NIQE	69.912	59.435	60.962	69.759	61.744	61.317	32.961
100602	NIQE	7.283	8.054	6.885	6.832	7.036	6.952	5.9664
100603	PSNR	14.584	18.765	17.148	20.580	18.396	18.012	23.613
100603	SSIM	0.562	0.690	0.584	0.701	0.628	0.617	0.867
100603	IL-NIQE	64.945	59.374	54.171	65.780	51.560	47.943	37.503
100603	NIQE	6.503	6.711	6.360	6.946	6.484	6.455	5.622
100605	PSNR	15.408	16.461	15.450	19.119	17.667	17.406	22.167
100605	SSIM	0.531	0.569	0.483	0.629	0.576	0.568	0.770
100605	IL-NIQE	91.358	81.761	81.389	85.474	78.275	73.778	51.919
100605	NIQE	7.819	7.652	6.863	6.792	8.183	7.738	5.4618