Enhancement of iterative back projection technique for super resolution image

Abstract

This study concerns in on improvements of the Iterative Back Projection (IBP) Super Resolution reconstruction technique. This technique suffers from self produces the ringing artefacts around the strong edge of the image and consumes a large number of iterations during the estimation process. Besides that, this technique has poor performances during handling the noisy image. In order to reduce the ringing artefacts, this thesis proposes a novel approach of the IBP reconstruction by hybrid with the edge enhancement regularization. First, the regulator uses the Sharp Infinite Symmetrical Exponential Filter (SISEF) to provide an accurate and precise edge guided during back projection to High Resolution (HR) grid. This precise edge guided prevents the IBP reconstruction from dislocating the HR projection. Thus, it reduces the ringing artefacts and increases the output quality averagely up to 1.755 dB. This SISEF regulator supplies accurate additional information and helps the estimator to save 70% iterations from conventional technique in determining the HR image. Second, in order to increase the robustness of IBP reconstruction, this study proposes an improvement with implementation of the Lorentzian Sharp Infinite Symmetrical Exponential Filter (LSISEF) regulations.