基于稀疏表示的Shearlet域SAR图像去噪.pdfVIP

第34 卷第9 期 电 子 与 信 息 学 报 Vol.34No.9 2012 年9 月 Journal of Electronics Information Technology Sept. 2012 基于稀疏表示的Shearlet 域SAR 图像去噪 刘帅奇* 胡绍海 肖 扬 (北京交通大学信息所 北京 100044) 摘 要：该文通过分析SAR 图像的噪声成因以及其斑点噪声模型，结合图像的稀疏表示理论提出一种基于稀疏表 示的Shearlet 域SAR 图像去噪算法。算法从整体上对SAR 图像进行去噪：首先对SAR 图像进行Shearlet 变换， 然后利用稀疏表示模型构造出去噪的最优化模型，在此基础上进行迭代去噪，然后重构SAR 图像得到去噪后的图 像。实验结果表明：该文所提出的算法不仅可以显著去除相干斑噪声，提高去噪图像的峰值信噪比(Peak Signal to Noise Ratio, PSNR)，还明显地改善了图像的视觉效果，更好地保留了图像纹理信息。 关键词：合成孔径雷达；图像去噪；稀疏表示；Shearlet 去噪；共轭梯度法 中图分类号：TN957.52 文献标识码： A 文章编号：1009-5896(2012)09-2110-06 DOI: 10.3724/SP.J.1146.2012.00200 Shearlet Domain SAR Image De-noising via Sparse Representation Liu Shuai-qi Hu Shao-hai Xiao Yang (Institute of Information Science, Beijing Jiaotong University, Beijing 100044, China) Abstract: After analyzing the causes of SAR image noise and speckle model, a SAR image de-noising method is presented in Shearlet domain from the theory of image sparse representation. The proposed algorithm is to de-noise SAR image from the entire image information: firstly, Shearlet transform is applied to the noise SAR image, then, the de-noised Shearlet coefficients are got based on iterative de-noising algorithm from noise optimization model which constructed by the model of sparse representation of the SAR image, finally, the clean SAR image is obtained from the de-nosing Shearlet coefficients. The experimental results show that the proposed algorithm can suppress speckle and improve the PSNR of de-noised image significantly, as well as improve visual effect of the image and retain the image texture information better. Key words: SAR; Image de-nosing; Sparse representat