A Super Resolution Algorithm Based On Content Regional Division

In the field of image process research, super-resolution algorithm is very important and widely used in practice. Some traditional super-resolution algorithms run fast and keep a consistent structure between the original image and super-resolution result, but cannot rebuild the detail information. Other deep learning algorithms can achieve better results, but are time-consume. We propose a new super-resolution algorithm based on content regional division. According to the similarity of image content, the image is divided into several parts. If patches of these areas can match with the patches in the database, the corresponding algorithm is used to replace them; otherwise, the self-similar method is used for super-resolution processing. The experiment shows that we can get acceptable high-resolution images at higher speed.