中文说明： 针对K-SVD算法和BM3D算法的不足，本文提出了基于字典学习和结构聚类的图像去噪算法。该算法首先通过字典学习得到含噪图像的冗余字典，然后对相似的图像块进行聚类构成块群，并通过迭代收缩和L1正则化约束，对同类的图像块在字典上进行稀疏表示，以达到降噪的目的。实验结果表明，在常规的图像处理上，本文提出的算法能较好的保留图像的结构信息，与K-SVD和BM3D等现有的流行算法相比，具有更高的峰值信噪比（PSNR）。
English Description：
Aiming at the shortcomings of K-SVD algorithm and BM3D algorithm, this paper proposes an image denoising algorithm based on dictionary learning and structural clustering. Firstly, the redundant dictionary of noisy image is obtained by dictionary learning, and then the similar image blocks are clustered to form a block group. By iterative shrinkage and L1 regularization constraints, the similar image blocks are sparsely represented in the dictionary to achieve the purpose of noise reduction. The experimental results show that the proposed algorithm can preserve the structure information of the image better in the conventional image processing, and has higher peak signal-to-noise ratio (PSNR) than the existing popular algorithms such as K-SVD and BM3D.