摘要
非下采样轮廓波变换(NSCT)是一种新的多尺度几何分析工具,具有平移不变性、多方向性和各向异性。与小波变换相比,NSCT能更好地表示图像中的边缘等信息。对合成孔径雷达图像进行NSCT分解,考虑其系数统计特性,基于BayesShrink对每个分解层的各个子带做多层阈值估计和软阈值收缩处理。实验结果表明,采用该方法得到的图像在视觉效果和客观衡量指标上均符合要求。
As a new multi-scale geometric analysis tool, Nonsubsampled Contourlet Transform(NSCT) has the characteristics of shift-invariance, multi-directionality and anisotropy. NSCT has the better representation of informations such as edges than wavelet transform. This paper decomposes Synthetic Aperture Radar(SAR) image by NSCT and considers its coefficients statistic characteristic. Based on BayesShrink, the multi-threshold estimation and the soft-threshold shrinkage in each subband of every decomposition layer are accomplished. Experimental results show that by using this method, the image meets the need of visual effect and objective measures.
出处
《计算机工程》
CAS
CSCD
北大核心
2010年第4期200-201,204,共3页
Computer Engineering
基金
国家自然科学基金资助项目(60802061
60702062)
河南省创新型科技人才队伍建设工程基金资助项目(084100510012)
河南省教育厅自然科学基金资助项目(2008B510001)
陕西省自然科学基金资助项目(2007F09)
关键词
合成孔径雷达
图像去噪
非下采样轮廓波变换
多阈值
Synthetic Aperture Radar(SAR)
image denoising
Nonsubsampled Contourlet Transform(NSCT)
multi-threshold