Abstract：Image Super-resolution Reconstruction (SR) aims to obtain high-resolution images from one or more low-resolution images. Since emerging in 1964, SR has been developing continuously and widely applied in different fields, from satellite imaging to medical image processing, to facial image analysis, video surveillance and biometrics. This survey retrospects the history of SR technique and provides a comprehensive overview of representative SR methods, with an emphasis on the recent work of deep learning-based approaches. We elaborate the details of various deep learning-based SR methods, including their strength and weakness, in terms of the deep learning model, architecture, message pass and network’s loss function. Finally, we discuss the possible research directions on SR technique.
