Deep Learning-Enabled Variational Optimization Method for Image Dehazing in Maritime Intelligent Transportation Systems

Table 1

SSIM comparisons (mean ± std) of various dehazed methods on all test images shown in Figure 5.

SSIM
		0.3	0.5	0.1	0.3	0.5	0.1	0.3	0.5
Haze	0.593 ± 0.106	0.580 ± 0.109	0.563 ± 0.111	0.746 ± 0.067	0.731 ± 0.073	0.713 ± 0.078	0.866 ± 0.038	0.853 ± 0.044	0.838 ± 0.049
DCP	0.680 ± 0.080	0.690 ± 0.079	0.646 ± 0.111	0.840 ± 0.052	0.814 ± 0.063	0.751 ± 0.081	0.897 ± 0.035	0.869 ± 0.042	0.829 ± 0.051
RIVD	0.637 ± 0.101	0.625 ± 0.105	0.607 ± 0.108	0.819 ± 0.055	0.839 ± 0.060	0.843 ± 0.065	0.839 ± 0.109	0.884 ± 0.022	0.901 ± 0.021
MSCNN	0.704 ± 0.079	0.690 ± 0.085	0.669 ± 0.090	0.913 ± 0.033	0.904 ± 0.042	0.887 ± 0.051	0.968 ± 0.018	0.967 ± 0.025	0.960 ± 0.032
Ours	0.796 ± 0.077	0.793 ± 0.083	0.779 ± 0.086	0.883 ± 0.054	0.898 ± 0.052	0.900 ± 0.054	0.884 ± 0.044	0.899 ± 0.042	0.908 ± 0.041