Image super-resolution based wavelet framework with gradient prior

Abstract

A novel super-resolution approach is presented. It is based on the local Lipschitz regularity of wavelet transform along scales to predict the new detailed coefficients and their gradients from the horizontal, vertical and diagonal directions after extrapolation. They form inputs of a synthesis wavelet filter to perform the undecimated inverse wavelet transform without registration error, to obtain the output image and its gradient map respectively. Finally, the gradient descent algorithm is applied to the output image combined with the newly generated gradient map. Experiments show that our method improves in both the objective evaluation of peak signal-to-noise ratio (PSNR) with the greatest improvement of 1.32 dB and the average of 0.56 dB, and the subjective evaluation in the edge pixels and even in the texture regions, compared to the "bicubic" interpolation algorithm. © 2011 Springer-Verlag.