多感受野特征与空谱注意力结合的高光谱图像超分辨率算法

曲海成, 王雅萱, 申磊

Hyperspectral super-resolution combining multi-receptive field features with spectral-spatial attention

QU Haicheng, WAND Yaxuan, SHEN Lei

表3 在Pavia Centre scene数据集上不同算法的对比结果

Tab.3 Comparison results of different algorithms on Pavia Centre scene dataset

算法 RMSE↓ MPSNR↑ CC↑ SAM↓ ERGAS↓ MSSIM↑ Bicubic 0.043 7 27.587 4 0.859 4 6.139 9 6.881 4 0.696 1 VDSR 0.041 9 27.882 1 0.865 9 6.700 4 6.699 1 0.724 2 RCAN 0.037 6 28.816 5 0.891 7 5.978 5 6.048 5① 0.771 9 TLCNN 0.043 1 27.668 2 0.856 3 6.901 3 6.913 9 0.714 1 3DCNN 0.039 6 28.411 4 0.881 3 5.866 9 6.266 5 0.750 1 DeepPrior 0.041 0 28.106 1 0.872 3 6.266 5 6.484 5 0.736 5 本文算法 0.028 1 31.942 2 0.940 6 6.557 7 6.742 3 0.878 6