基于图像非局部自相似性与分类字典学习的超分辨率重建算法

Image Super-resolution Reconstruction Algorithms Based on Nonlocal Self-similarities and Classification Dictionaries Learning

提出一种新的图像稀疏表示方法,该方法自适应地利用图像的局部与非局部冗余信息,根据图像的非局部自相似性,构造出一个非局部自回归模型,将其作为数值保真项.利用主成分分析方法及高分辨率的样本图像块学习构建紧凑的分类字典,通过限制迭代次数用以减少字典训练的计算量,同时字典在稀疏域中能够自适应选取.实验结果表明,与其他几种基于学习的算法比较,本文算法无论是在峰值信噪比、结构相似性上还是主观视觉效果上都有显著提高.

In this paper,we propose a novel image sparse representation method which exploits adaptively the local and nonlocal redundancies of the image. According to the nonlocal self-similarity of the image,a nonlocal autoregressive model is proposed and taken as the data fidelity term in sparse representation. The principal component analysis（ PCA） technique and high-quality example image patches learning are used to construct the classification dictionaries. We update the dictionaries in several iterations to reduce computational cost. At the same time,the dictionaries are selected in sparse domain adaptively. Extensive experiments on single image validate that the proposed method,compared with several other state-of-the-art learning based algorithms,achieves much better results in terms of both peak signal-to-noise ratio（ PSNR） and structural similarity（ SSIM） and subjective visual perception.