Real time super resolution image reconstruction

The super-resolution (SR) imaging is a process of constructing high resolution image using several shifted low resolution images of the same scene. The process should ideally be fast, and should add sharpness and detail, both at edges and in regions without adding artifacts. SR imaging works in two major phases: registration and interpolation. Existing registration techniques are computationally time consuming. Current interpolation methods suffer from blocking artifacts and also smoothens the edges and is unable to restore high frequency details in image. In this paper we propose a technique to yield a real time color super resolved image from four low resolution color images. We propose computationally efficient registration and robust interpolation for SR image reconstruction. Our technique is computationally fast having clear perspective of real time implementation. Experiments with several test images show the effectiveness of proposed technique.