Single-Frame Image Super-Resolution Based on Local Similarity Projection Model

Abstract

Local geometric similarity model can be edge preserving because of using self-similarity. But its interpolation weights must be solved by inverse linear equation. Based on the model, we proposed a new image super-resolution algorithm. It interpolate in the quincunx lattice to automatically capture edge characteristics in the quincunx lattice, then the local similarities between the original and interpolated pixels are enhanced by projection. The experimental results show that our method can reconstruct the high resolution image with edge-preserving.