Super resolution image reconstruction using total variation regularization and learning-based method

Super resolution is not only a key word in active research but also has become a sale point for the recent consumer product such as HDTV. Among a lot of proposals for super resolution image reconstruction, the total variation regularization (TV) method seems to be the most successful approach with sharp edge preservation and no artifacts. The TV regularization method still has two problems. One is the computational time and the other is the texture interpolation. In this paper, we propose a system which offers solutions to these problems. In our system, the number of TV regularization processes is reduced compared with the conventional method, and the learning-based method is introduced instead of the texture interpolation. The experimental result shows that our approach realizes a super resolution image reconstruction with high performance and short computational time.