结合图像的局部相关性及非局部相似性的多尺度分块压缩感知  

Multiscale block compressed sensing by combing image local correlation and nonlocal similarity

作　　者：陈书贞[1] 李光耀[1] 练秋生[1] 

机构地区：[1]燕山大学信息科学与工程学院,河北秦皇岛066004

出　　处：《燕山大学学报》2013年第6期547-553,共7页Journal of Yanshan University

基　　金：河北省自然科学基金资助项目(F2010001294)

摘　　要：标准的图像压缩感知算法未利用像素间的邻域结构信息和图像子块的自相似性。针对这一问题,本文将图像分成重叠的图像子块,用冗余字典自适应地稀疏表示图像,同时将用自回归模型表示的图像局部相关性和非局部相似性作为先验知识运用到压缩感知图像重构中,提出了结合图像的局部相关性和非局部相似性的多尺度分块压缩感知方法。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。The standard image compressed sensing ignores the neighborhood structure information of the image pixels and the similarity of image patches. To address this issue,the whole image was cut into overlapping patches and the image patches are represented by adaptive redundant dictionary.The image local correlation which is represented by autoregressive model and image nonlocal similarity was used in image compressed sensing as prior.And then a new multiscale block compressed sensing algorithm by combing image local correlation and nonlocal similarity is proposed in this paper.Simulation results show that the performance of the proposed algorithm has significant performance improvement in visual quality of the reconstructed image and peak signal-to-noise ratio.

关 键 词：多尺度分块压缩感知 稀疏表示 图像重构 局部相关性 非局部相似性 

分 类 号：TN911.73[电子电信—通信与信息系统]