Adaptive super resolution algorithm based on RBM dictionary learning

According to the SCSR (sparse coding sparse representation), algorithm based on general dictionary can not characterize the various structural types of image and global sparse reconstruction introduced these 2 shortcomings of redundant, proposed super-resolution algorithm adaptive decomposition based on MCA (morphological component analysis). This algorithm, first of all, using sparse K-SVD method to obtain the low resolution Dictionary of training for low resolution image reconstruction and down sampling as dictionary training samples, improved the correlation between low resolution images and reconstructed the dictionary. Secondly, in the reconstruction phase, the MCA method is used to extract the texture components of the image to reconstruct the sparse image. Experimental results show that compared with other advanced algorithms, the proposed algorithm is able to recover the image edge details better, and the reconstructed quality is better.