Image Super-resolution Using Mid-level Representations

Li Yang, Yaxing Wang, Xiaomin Mu, Yaping Wang

Abstract


An end-to-end six layers convolutional neural network(CNNs) structure is proposed to realize single image super-resolution reconstruction. According to the study of image Mid-Level Representations(MLR), the network layers are divided into two parts. The first part is used to obtain shallow informations by transferring features, and the second part is used to realize the feature enhancement. The input of the network is low-resolution (LR) image, and the output is the superresolution (SR) image. Promising experimental results are obtained with higher precision.

Keywords


Convolutional neural network, Super-resolution reconstruction, Mid-level image representations, Parameters transfer

Publication Date


2016-11-18 00:00:00


DOI
10.12783/dtetr/iect2016/3750

Refbacks

  • There are currently no refbacks.