Fig. 3
The architecture of multi-scale attention block (MSAB). Our MSAB mainly consists of 3 attention blocks (AB) with different scales of convolution kernel to explore the multi-scale features for the better enhancement.
Fig. 3 | Scientific Reports
