Learning based image super resolution using sparse online greedy support vector regression

Abstract

Kernel filters with sparse solutions has become highly advantageous because of the reduced computation. Sparse online greedy support vector regression algorithm, computes the coefficients only after the generation of sparse dictionary. This paper super resolves a low resolution image to high resolution image, with the model generated from the training set using sparse online greedy support vector regression. The method is evaluated with super resolution using support vector regression. Comparisons are done on the PSNR, time and memory scales. The sparse online greedy support vector regression shows good improvement in these scales.