Double Sparse Dictionary Learning for Image Super Resolution

This paper presents a novel double sparse dictionary learning method for the image upscaling which aim is to recover a high resolution image based on a given low resolution input. The proposed algorithm is motivated by sparse signal representation and compressive sensing theory. We incorporate the double sparse dictionary into sparse representation model. In dictionary learning phase, we impose l1-norm not only on coefficient but on dictionary as well. Since the double sparse dictionary reduces the coherence between observation matrix and dictionary, it is stable under noise and can accurately recover the original signal form its measurement. Experimental results on benchmark image data set are presented and compared with some exiting super resolution method. The result demonstrates the advantages of the proposed method.