Image super resolution by sparse linear regression and iterative back projection

Abstract

This paper presents a method which focus on the increase of visual quality of SR image reconstructed from input low resolution image. Similar to the framework as exploited in [1], a modified algorithm is developed which is based on sparse linear regression and iterative back projection. Different from the techniques used in [1] [6], a feature sign search algorithm [17] is used to find the relevant features of the regression function under a priori assumption. Furthermore, a modified Gaussian high pass filter is additionally used for the refinement of the initial reconstructed SR image through iterative back-projection technique to reduce visual artifacts. Experimental results conclude that this modified approach achieves better quality of reconstructed SR images than the other similar SR methods.