Table 3.

Quantitative comparison of GTMNet with other methods on SateHaze1k dataset.

Setting	Thin fog		Moderate fog		Thick fog		Params
	PSNR/dB	SSIM	PSNR/dB	SSIM	PSNR/dB	SSIM
DCP4	13.15	0.725	9.78	0.574	10.25	0.585	–
DehazeNet5	19.75	0.895	18.12	0.855	14.33	0.706	0.009 M
Huang (SAR)7	24.16	0.906	25.31	0.926	25.07	0.864	–
SkyGAN8	25.38	0.925	25.58	0.904	23.43	0.893	–
TransRA14	25.20	0.930	26.50	0.947	22.73	0.875	–
Dehazeformer-T22	24.51	0.968	26.38	0.969	22.02	0.932	0.686 M
Dehazeformer-S22	24.60	0.968	26.52	0.971	22.11	0.933	1.283 M
Dehazeformer-B22	25.18	0.971	26.74	0.973	22.23	0.934	2.514 M
Dehazeformer-L22	25.89	0.974	27.23	0.973	23.14	0.942	25.44 M
GTMNet-T	25.14	0.970	26.73	0.971	22.31	0.935	0.798 M
GTMNet-S	24.72	0.967	26.61	0.972	22.29	0.935	1.396 M
GTMNet-B	24.89	0.970	27.22	0.973	23.02	0.939	2.629 M
GTMNet-L	25.52	0.972	27.18	0.973	22.96	0.940	25.87 M