Fast single-image upsampling with relative edge growth rate priors
Chang Su, Tao Li · 2015
In this paper, we propose a fast image up-sample algorithm using the priors of the spatial-frequency (SF) changes between high-resolution (HR) images and their low-resolution (LR) counterparts. A novel feature, the relative edge growth rate (REGR), is proposed to represent the SF of an image. The SF priors are thus described by REGR and learned from real-world images. With the priors, a visual pleased HR image can be quickly estimated from the interpolated image of a LR image by recovering its SF. Experimental results show that the proposed algorithm significantly improves the visual quality of the estimated HR images.