基于非局部相似性和交替迭代优化算法的图像压缩感知.pdfVIP

第 28 卷第2 期 Vol. 28 No.2 信号处理 2012 年2 月 SIGNAL PROCESSING Feb. 2012 基于非局部相似性和交替迭代优化 算法的图像压缩感知 陈书贞李光耀练秋生 (燕山大学信息科学与工程学院，秦皇岛 066仪)4) 摘 要:压缩感知理论突破了信号带宽对奈奎斯特采样定理的限制，并且实现了在数据采样的同时进行压缩。目 前压缩感知系统通常利用图像在某个变换域具有稀疏性的先验知识，从少量观测值中重构原始图像。本文利用 图像像素的邻域结构信息及图像子块的相似性，将图像的非局部相似性作为先验知识运用到压缩感知图像重构 中。结合图像的非局部相似性及其在变换域的稀疏性先验知识，提出了基于非局部相似性和交替选代优化算法 的图像压缩感知重构算法，该算法利用迭代阙值法和非局部全变差来交替选代求解变换域的稀疏性优化问题和 非局部相似性的优化问题。实验结果表明，本文算法可以有效提高图像重构的视觉效果和峰值信噪比。 关键词:压缩感知:稀疏袭示:图像重构;非局部相似性 中固分类号: TN91 1. 73 文献标识码:A 文章编号: 1∞3-0530(2012)02-02∞-俑 Image compressed sensing based on nonlocal similarity and alternating iterative optimization algorithm . CHEN Shu-zhen U Guang-yao UAN Qiu-sheng (Institute of Infonnation science and technology , Yanshan university , Qinhuangdao 066(剧， China) Abstract: Compressed Sensing Ïs a new technique for simultaneous data sampling and compression. It breaks through 由e limits to Ny甲ist sampling theorem which needs very wide signal bandwidth when sampling. Currenùy ，∞mpressed 配，nsi吨 system used 由e prior 阳t 由e image has sparsity in some transfonn domain to reconstruct the original image from fewer me画· 田ements. In this paper , the nonlocal similarity was used in image compressed sensing and combined wi出由e sparsity as pri- or. Hence , the neighborhood structure information of 出.e image pixels and the similarity of images are fully used. On the bàSis of the nonlocal similarity prior and the im且在;e has sparsity in some transfonn domain , a new image compressed sensing algo-