Lightweight refined networks for single image super-resolution

Recently, software architectures applied to physical agents have become a boost from the emerging Artificial Intelligence(AI). In these smart physical agents, simultaneously image information processing appears vital in particular. Single Image Super Resolution(SISR) serves as the foundation of the image process, presenting its prospects driven by deep learning(DL) methods. In these DL methods, convolutional layers are stacked to implement a mapping between the original low resolution(LR) image and the high resolution(HR) image. Despite improved performances, convolution neural networks(CNN) based methods with large parameters are so complex and time-consuming, which is difficult to apply in mobile devices. To tackle the above issue, we propose Distillation Information Block(DIB) and Feature Information Refined Block(FIRB). In our proposed Lightweight Refined Networks for single image Super-Resolution(LRSR), to lengthen our network with fewer parameters increased, the proposed DIB grasps more information by using a large receptive field. To enhance the information utilization, we build FIRB to refine advanced features and recover more details. Furthermore, with the compact structure, the execution time can be comparatively reduced with higher performance. We conduct extensive experiments on different datasets, which demonstrate that the proposed method performs comparatively better compared with the state-of-the-art lightweight methods.