基于字典学习和非局部相似的超分辨率重建

为提高单帧降质图像的分辨率,提出了一种基于字典学习和非局部相似性的超分辨率重建算法。该算法主要将高分辨率图像减去利用迭代反投影重建结果得到差值图像,再利用K-奇异值分解(K-SVD)算法和联合字典生成的思想形成的字典训练方法,训练差值图像块和低分辨率图像块得到对应的高、低分辨率字典用于超分辨重建。此外,引入非局部相似性的正则项约束以提高重建图像的质量。实验结果表明,所提算法重建得到的图像在主观视觉效果和客观评价上优于基于例子学习的超分辨率算法。

Super-resolution reconstruction based on dictionary learning and non-local similarity

To deal with the single-image scale-up problem, a super-resolution reconstruction algorithm based on dictionary learning and non-local similarity was proposed. The difference images between the high-resolution images and results of using iterative back-projection image reconstruction were obtained, and then the high and corresponding low dictionaries could be co-generated by training difference image patches and the corresponding low-resolution image patches via using K-Singular Value Decomposition (K-SVD) algorithm which was combined with the idea that the high and low dictionaries could be co-trained for super-resolution reconstruction. In addition, a non-local similarity regularization constraint was introduced in the new algorithm to further improve the quality of the reconstructed images. The experimental results show that the proposed algorithm achieves better results than learning-based algorithms in terms of both visual perception and objective evaluation.