Image Magnification through Non-local Edge Directed Interpolation

Abstract

In this paper, we present a non-local edge directed interpolation method to magnify single frame image. The basic idea is to first extract the edge features and geometric features of image patches and use these feature matrices to match with the image patches to be interpolated employing the SSIM algorithm. Then taking advantage of the geometric duality between the low-resolution covariance and the high-resolution covariance to obtain the optimization coefficients and exploit multiple linear regression model (MLR) to reconstruct the image patch. The new algorithm can preserve the edge features and natural appearance of images efficiently. Simulation results demonstrate that our new interpolation algorithm substantially improves the subjective and objective quality.