Auto coder-decoder (CODEC) model based sparse representation for image super resolution

Abstract

In our daily life, the high quality image is widely used in varieties of fields, but sometimes we cannot capture the image with idea resolution due to some influences. For solving the resolution limitation of imaging sensors, the image super resolution (SR) representation technology is widely researched. Considering the advantage of sparse representation, the dictionary learning based methods is widely studied. However, landmark atoms cannot provide the representations of images, since the general feature extractors is universally applicable in feature extraction. To overcome the drawbacks, an auto coder-decoder (CODEC) model is proposed to extract representative features from low resolution (LR) images. The experimental results indicate the proposed method can obtain better effect than other methods.