Super-resolution Reconstruction Based on Geometric Dictionary Learning and Coupled Regularization
MO Jian-we · Infrared Technology · 2015
Traditional super-resolution algorithms based on sparse representation of image patches exploit single redundant dictionary to represent the image patches that contain various textures, which can not reflect the differences of various image patches types. In order to overcome this disadvantage, this paper proposes a single image super resolution reconstruction method based on dictionary learning and coupled regularization, by exploring the local property of image patches. A large number of training image patches are clustered into several groups by their property, from which the corresponding geometric are learned via K-SVD algorithm which is combined with the idea that the high and low resolution dictionaries can be co-trained. In addition, a coupled regularization of local steering kernel regression and non-local similarity is introduced into the proposed method to further improve the quality of the reconstructed images. Experiment results show that the proposed method both increases the evaluation parameters and improves the visual quality of the edges and the details significantly.