基于Shearlet的双正则化图像压缩采样恢复.pdfVIP

Computer Engineering and Applications 计算机工程与应用 2014 ，50（11） 145 ⦾图形图像处理⦾ 基于Shearlet 的双正则化图像压缩采样恢复 1 2 许志良 ，况庆强 XU Zhiliang , KUANG Qingqiang2 1.深圳信息职业技术学院 软件学院，广东 深圳 518172 2.江西师范大学 物理与通信电子学院 江西省光电子与通信重点实验室，南昌 330022 1.Department of Software, Shenzhen Institute of Information Technology, Shenzhen, Guangdong 518172, China 2.Key Laboratory of Photoelectronics Telecommunication of Jiangxi Province, College of Physics Communication Electronics, Jiangxi Normal University, Nanchang 330022, China XU Zhiliang, KUANG Qingqiang. Shearlet based hybrid regularizers for image compressive sampling reconstruction. Computer Engineering and Applications, 2014, 50（11）：145-148. Abstract ：The orthogonal wavelet fails to provide an optimal sparse representation for images that contain texture details due to limited direction, and the current regularization method is singleness. In this paper, it proposes a compressed sensing reconstruction algorithm based on Shearlet sparse representation and compound regularizers, the algorithm uses Shearlet for image sparse representation, and the problem is solved by alternating minimization algorithm. The experimental results indicate that the visual quality and PSNR of reconstructed image is improved by proposed algorithm compared with single total variation regularizer method and wavelet transform. Key words ：compressive sampling; Shearlet transform; regularizer 摘 要：针对图像压缩采样中正交小波变换方向有限和单一正则化的问题，提出了一种基于Shearlet 的双正则化图 像压缩采样恢复算法。该算法用Shearlet 作为图像的稀疏表示，用交替最小化对联合正则化模型进行求解。实验结 果表明，该算法恢复的图像与单一的全变分正则化方法和小波变换相比有更好的视觉效果，更高的峰值信噪比。 关键词：压缩采样；剪切波变换；正则化 文献标志码：A 中图分类号：TN911.73 doi ：10.3778/j.issn.1002-8331.1309-0285 1 引言 值对原始图像进行重构。目前主要采用二维的正交小 传统的数据获取与重构遵循奈奎斯特采样定理，即 波对图像进行稀疏表示，利用合适的重构算法对含有正 精确地重构原始信号需要至少以原始信号两倍带宽的 则化项的重构模型进行求解来重构原始图像[ 8 。但对 抽样率对信