End-to-End Super-Resolution for Remote-Sensing Images Using an Improved Multi-Scale Residual Network

Remote-sensing images constitute an important means of obtaining geographic information. Image super-resolution reconstruction techniques are effective methods improving the spatial resolution remote-sensing images. Super-resolution networks mainly improve model performance by increasing network depth. However, blindly depth can easily lead to gradient disappearance or explosion, difficulty training. This report proposes a new pyramidal multi-scale residual (PMSRN) that uses hierarchical residual-like connections and dilation convolution form block (MSDRB). The MSDRB enhances ability detect context information fuses features through feature fusion structure. Finally, complementary global local is added structure alleviate problem useful original ignored. experimental results showed that, compared with basic network, PMSRN increased peak signal-to-noise ratio up 0.44 dB structural similarity 0.9776.