A Complete Imaging System with A Novel Image Super-Resolution Method

Single image super-resolution (SR) has been widely studied in recent years as a crucial technique for many applications. This paper proposes a complete imaging system with a novel image super-resolution method based on convolutional neural network (CNN). Different from the previous papers, the novelty of this paper mainly reflects from two aspects. First, the whole architecture of the imaging system is showed and illustrated. Secondly, a novel network model for SR is built, which contains deconvolution layer and residual network. Results validate that the proposed method performs well in terms of both the objective evaluation and the subjective perspective.