Image super-resolution method based on non-local means and self similarity

In this paper, we propose an image super-resolution method based on the non-local means and the self similarity. Various super-resolution methods can correctly estimate the missing high frequency components of enlarged images. However, they mostly require high computational costs, which is not suitable for real-time processing. For a super-resolution with low computational costs, the proposed method is simply realized via the block matching technique with a small search area. Since it utilizes the image self similarity and sparsity, it produces visually efficient interpolated images. In the simulation, it is shown that the proposed method greatly outperforms the bicubic in a visual quality of enlarged images, objectively and perceptually.