基于多尺度特征映射网络的图像超分辨率重建

段然,周登文,赵丽娟,柴晓亮

Image super-resolution reconstruction based on multi-scale feature mapping network

Ran DUAN,Deng-wen ZHOU,Li-juan ZHAO,Xiao-liang CHAI

表 1 4倍放大因子下与其他5种SR方法在Set5、Set14和BSD100上的平均PSNR和SSIM对比

Tab.1 Average PSNRs and SSIMs compared with other SR methods on Set5，Set14 and BSD100 （4×）

SR方法 PSNR/dB, SSIM Set5 Set14 BSD100 平均值 双三次插值 28.43, 0.810 25.77, 0.703 25.98, 0.671 26.73, 0.728 A+[26] 30.30, 0.859 27.38, 0.752 26.82, 0.710 28.17, 0.774 SRCNN[6] 30.49, 0.863 27.50, 0.751 26.91, 0.712 28.30, 0.775 FSRCNN[8] 30.71, 0.865 27.60, 0.753 26.97, 0.713 28.43, 0.777 文献[10]方法 30.80, 0.873 28.60, 0.792 26.96, 0.711 28.78, 0.792 本文方法（MSE） 31.32, 0.881 27.78, 0.773 27.40, 0.736 28.83, 0.797