A Distributed Fusion Framework of Multispectral and Panchromatic Images Based on Residual Network

Remote sensing images have been widely applied in various industries; nevertheless, the resolution of such is relatively low. Panchromatic sharpening (pan-sharpening) a research focus image fusion domain remote sensing. Pan-sharpening used to generate high-resolution multispectral (HRMS) making full use low-resolution (LRMS) and panchromatic (PAN) images. Traditional pan-sharpening has problems spectral distortion, ringing effect, low resolution. The convolutional neural network (CNN) gradually pan-sharpening. Aiming at aforementioned problems, we propose distributed framework based on residual CNN (RCNN), namely, RDFNet, which realizes data three channels. It can make most information spatial LRMS PAN proposed employs architecture best outcome previous step channel, so that subsequent acquires much more information. Moreover, two feature extraction channels are extract features MS respectively, using module, different scales for channel. In this way, distortion loss reduced. Employing from four satellites compare results experiment show RDFNet superior performance improving preserving information, good robustness generalization quality.