基于不变学习的真实雾霾去除方法

孟小哲,冯钰新,苏卓,周凡

Real-world dehazing method with invariant learning

Xiaozhe MENG,Yuxin FENG,Zhuo SU,Fan ZHOU

表 1 不同方法在RTTS数据集上去雾结果的无参考图像质量评价

Tab.1 No-reference image quality evaluation of different dehazing results methods on RTTS dataset

方法 NIMA[34] BIQME[32] PM2.5[33] 输入 4.325 0.504 198.250 MSBDN[19] 4.140 0.548 166.783 FFA[6] 3.718 0.524 189.193 DehazeFlow[5] 4.687 0.530 177.369 SGID[3] 3.694 0.545 178.359 DAD[23] 4.005 0.561 108.403 D4[31] 4.631 0.555 163.386 IDE[21] 4.586 0.553 150.240 PSD[22] 4.345 0.514 151.900 SLA[24] 4.598 0.524 138.242 本文方法 4.703 0.577 137.436