Transfer Learning Based on A+ for Image Super-Resolution

摘要: Example learning-based super-resolution (SR) methods are effective to generate a high-resolution (HR) image from single low-resolution (LR) input. And these SR have shown great potential for many practical applications. Unfortunately, most of popular example approaches extract features limited training images. These images insufficient super resolution task. Our work is transfer some supplemental information other domains. Therefore, in this paper, new algorithm Transfer Learning based on A+ (TLA) proposed First, we datasets construct dictionary. Then, sample selection, more samples supplemented the basic samples. In experiments, seek explore what types can provide appropriate Experimental results indicate that our approach superior when transferring containing similar content with original data.