基于稀疏表示的图像超分辨率重构算法

为提高单幅图像的分辨率，提出一种基于稀疏表示的图像超分辨率重构方法。该方法的核心是联合训练高分辨率和低分辨率字典，然后利用所得字典求解高、低分辨率下图像块共有的稀疏表示系数。与已有的基于稀疏表示的图像超分辨重构算法相比，该算法在求解稀疏表示系数时并未采用拉格朗日乘子将稀疏度和重构误差相结合，而是利用对偶模型求解原始的带约束优化问题。实验表明，与其他图像超分辨率重构方法相比，该方法所需手动调节参数较少，重构效果较好。

Image Super- Resolution Reconstruction Algorithm Based on Sparse Representation

In order to enhance the resolution of single degraded images, a method of super - resolu- tion reconstruction is proposed via sparse representation in this paper. The core of this method is to train high- and low -resolution dic-tionaries jointly and employ a set （ff common sparse coefficients to represent high - and low -resolution images under corresponding dictionaries. Compared with the existing image super- resolution reconstruction algorithms based on sparse representation, the proposed method establishes dual models to solve the original optimization problem in the stage （ff sparse coding instead of combining sparsity and reconstruction error by Lagrange multipliers. Experimental results demonstrate that, compared with other approaches, a less number of parameters are manually set up, the reconstruction results of this method are also improved.