Single text image super-resolution based on edge-compensated autoregressive model

Text image super-resolution technique is widely used to improve the text image quality in different resolution without distortion. However, the existed algorithms for text image super-resolution fail to work under many conditions, such as image blurring, edge discontinuity, edge artifact, etc. In this paper, we propose a novel algorithm framework to solve these above problems, in which we firstly obtain a high resolution text image with a sharper edge by using edge map as the priori condition, and then the edge signals of high-resolution image compensating edge residual is computed by iterative back-projection, finally the false edges and sharpen edges are suppressed at the same time. Experimental results demonstrate the effectiveness of our proposed algorithm compared to previously reported methods.