摘要
针对SAR图像统计特性,构建具有良好的平移不变性、方向选择性和自适应近似最优匹配性能的复Bandelet基函数,并提出一种相干斑抑制算法.该算法基于全变差构建复Bandelet寻优的目标函数;采用广义交叉验证准则,在不需要估计噪声方差的情况下,自适应获取各个分解层的渐进最优阈值.实验与经典空域滤波、基于小波、双树复小波和Bandelets的滤波方法进行了比较,结果表明基于复Bandelets的滤波方法在有效滤波的同时更好地保护了图像的几何结构信息,客观衡量指标和视觉效果都有较明显的改善.
Considering statistic characteristics of synthetic aperture radar（SAR） images,a novel basis named complex bandelet（CB） was presented and applied on SAR image speckle reduction.The CB has characteristics of good shift-invariance,multi-directionality and self-adaptively approximate optimal matching.Cost function of CB searching optimization was constructed based on total variation（TV）.And then generalized cross validation（GCV） was introduced to get asymptotic optimal thresholds without estimating noise variance.Numerical tests show that CB-based filter provided improvements both in visual effects and quantitative analysis, which are embodied in smoothness of uniform fields, continuity and clearness of geometric structure information simultaneously. The compared filters are classical spatial domain filters, wavelet Iransform-based filter, dual-tree complex wavelet transformbased filter, and bandelet transform-based filter.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2009年第9期1880-1884,共5页
Acta Electronica Sinica
基金
国家自然科学基金(No.60802061
60702062)
河南省创新型科技人才队伍建设工程(No.084100510012)
河南省教育厅自然科学基金(No.2008B510001)
陕西省基金(No.2007F09)
作者简介
杨晓慧 女，1978年生于河南许昌．河南大学副教授，硕士生导师．研究方向为多尺度几何分析优化设计及其应用等．E-mail：xhyang@henu．edu．cn
焦李成 男，1959年生于陕西白水．西安电子科技大学教授，博士生导师，IEEE高级会员．研究方向为非线性科学、智能信息处理、小波理论及应用和机器学习等．
