多尺度稀疏表示的Contourlet域图像去噪  

Image Denoising Based on Multiscale Sparse Representation in Contourlet Domain

作　　者：刘小妹[1,2] 夏平[1,2] 雷帮军[1] 

机构地区：[1]三峡大学智能视觉与图像信息研究所,湖北宜昌443002 [2]三峡大学计算机与信息学院,湖北宜昌443002

出　　处：《三峡大学学报（自然科学版）》2014年第6期108-112,共5页Journal of China Three Gorges University：Natural Sciences

基　　金：国家自然科学基金项目(61272237);楚天学者科研基金项目(KJ2012B001)

摘　　要：提出了Contourlet域多尺度稀疏表示的自适应阈值图像去噪算法.首先,在分析了Contourlet域多尺度图像稀疏表示的基础上,提出了Contourlet域自适应阈值去噪算法;其次,详细地讨论了该算法选择不同的Contourlet域参数对图像去噪性能的影响,并比较了在等同的条件下Contourlet域与Wavelet域的去噪效果.实验结果表明,本文算法以及选择合适的Contourlet域参数进行图像去噪能有效地抑制图像噪声、保留边缘和轮廓信息.An improved hard threshold denoising algorithm of Contourlet domain is proposed. Discussed in detail the performance impact of image denoising because of the selection of Contourlet domain parameters in the algorithm. Firstly, the advantages of multiscale geometric analysis are described. Secondly, proposed an amendments hard threshold denoising algorithm based on Contourlet transform;Finally, discussed in detail the performance impact of image denoising due to select a different Contourlet domain parameter in the algo- rithm, and compared with the effect of Contourlet domain denoising and wavelet domain denoising in the same decomposition. Experimental results show that it is essential to select the appropriate Contourlet domain pa- rameters for image denoising; meanwhile, Contourlet domain for noise reduction can better suppress the im- age noise and keep the edge and contour effectively.

关 键 词：图像去噪 CONTOURLET域 多尺度稀疏表示 小波变换 阈值 

分 类 号：TN391.9[电子电信—物理电子学]