Efficient Reconstruction of Super-Resolution Images using an Improved Orthogonal Matching Pursuit Algorithm

This paper presents an improved reconstruction algorithm of super-resolution images using the compressed sensing theory. In particular, Gaussian low-pass filters are proposed to reduce the coherence between the observation matrix and transform matrix for signal representation such that the problem of super-resolution can be formulated as a proper CS reconstruction problem. Experimental results have shown that the proposed algorithm can reduce the computational cost for super-resolution reconstruction from a single image without loss of subjective visual quality.