Journal of Remote Sensing / 2021 / Article / Tab 2

Research Article

Multisensor Remote Sensing Imagery Super-Resolution with Conditional GAN

Table 2

Performance comparisons with state-of-the-art algorithms in terms of 4 evaluation metrics on simulated dataset and the introduced OLI2MSI dataset. Note that a higher value of PSNR and SSIM is better, while for LPIPS and NIQE score, a lower value is better.

AlgorithmsTrained on simulated datasetTrained on OLI2MSI dataset
PSNRSSIMLPIPSNIQEPSNRSSIMLPIPSNIQE

Bicubic33.240.8790.35122.2433.240.8790.35122.24
SRCNN33.660.8960.23718.1235.260.9070.21818.46
ESPCN33.650.8950.22818.1735.410.9100.21218.05
VDSR33.640.8950.21517.8635.870.9190.18116.61
SRResNet33.590.8960.21717.5537.130.9390.13113.25
EDSR33.560.8950.21216.7637.680.9460.11312.08
DBPN33.570.8950.21417.1137.600.9450.11612.26
RCAN33.580.8960.21517.0537.850.9470.11111.86
DRN33.600.8960.21717.3137.540.9440.11912.40
SRGAN32.700.8780.1456.0633.630.8790.0812.75
ESRGAN32.320.8780.1515.9033.820.8880.0752.45
cDRSRGAN(ours)33.590.8960.20416.8337.410.9430.0579.81