High Frequency Super-Resolution for Image Enhancement

Bayesian based MF-SR (multi-frame super-resolution) has been used as a popular and effective SR model. However, texture region is not reconstructed sufficiently because it works on the spatial domain. In this paper, we extend the MF-SR method to operate on the frequency domain for the improvement of HF information as much as possible. For this, we propose a spatially weighted bilateral total variation model as a regularization term for Bayesian estimation. Experimental results show that the proposed method can recover texture region with reduced noise, compared to conventional methods.