基于Contourlet域隐马尔可夫树模型的SAR图像滤波方法

A Method for De-speckling of SAR Image Based on CHMT Model

为了获得更好的合成孔径雷达(SAR)图像滤波效果,提出一种基于Contourlet域隐马尔可夫树(CHMT)模型的SAR图像滤波算法.提出基于粗分类的系数绑定方法,提高了CHMT模型参数的解算速度;综合应用对数变换、循环平移和均值校正等方法,建立了针对SAR图像乘性斑点噪声模型的统一滤波处理框架,并将基于CHMT模型的滤波算法融入该框架之中;通过对SAR影像进行滤波实验,并将该滤波算法与Lee滤波、小波软阈值滤波等方法进行了比较.可视效果和统计指标显示:基于粗分类的系数绑定方法在改善滤波效果的同时,对CHMT模型解算的速度有很大的提高;在统一滤波框架下,基于CHMT方法的滤波效果优于其他的几种滤波方法.

A synthetic aperture radar（SAR） image filter based on Contourlet-domain hidden Markov tree （CHMT） is proposed to achieve better result of de-speckled SAR image. First of all, the coarse-classification-based tying method for contourlet coefficients is designed to speed up the parameters estimation; and then, by working together with the LOG transform, mean rectification and cycle-spin, a general SAR image de-speckling workflow, in which the coarse- classification-based tying method for CHMT is applied to de- noise the simulated image and the true SAR image, is generated; at last, the results are compared to those of Leefilter , Wavelet soft threshold filter and other commonly-used filters. The visual effects and the statistical parameters indicate that the coarse-classification based tying method for CHMT is much faster than the other tying methods, and the CHMT based de-speckling method for SAR image can achieve better result than some commonly-used filters.