摘要
为了反映合成孔径雷达图像中斑点噪声尖峰厚尾的统计特征,使用拖尾Rayleigh分布来描述斑点噪声.基于Gamma先验分布和斑点噪声的拖尾Rayleigh分布,推导出了合成孔径雷达图像的最大后验概率滤波方程,并给出了它在特定特征参数时的解析形式.使用Mellin变换从观察图像估计拖尾Rayleigh分布的未知参数.给出了在斑点噪声的拖尾Rayleigh分布下的最大后验概率降噪试验和量化指标.为了消除滑动窗大小和噪声强度对降噪结果的影响,给出了降噪能力随滑动窗大小和噪声方差的动态变化关系.结果表明,拖尾Rayleigh分布尖峰厚尾的特征符合斑点噪声的真实统计特性,因此与Rayleigh分布以及Kuan滤波相比,基于斑点噪声的拖尾Rayleigh分布的最大后验概率滤波具有较强的降噪能力.
In order to reflect the statistics of high peak and heavy tail, speckle in synthetic aperture radar images is modeled as heavy- tailed Rayleigh distribution. First, based on Gamma prior distribution and heavy-tailed Rayleigh distribution of speckle, the maximum a pesteriori filtering equation is proposed and its analytical form is provided in given characteristic parameter. Second, parameters of heavy-tailed Rayleigh distribution are estimated from the observed image using Mellin transformation. Last, maximum a pesteriori de-speckling experiments and their quantitative measures are given. In order to eliminate the influence of window size and noise intensity on de-speckling results, dynamic relations of the de-speckling capability to noise variance and window size are suggested respectively. Results demonstrate that the heavy-tailed Rayleigh distribution accords with the real statistics of speckle, so the maximum a pesteriori filter in heavy-tailed Rayleigh distribution of speckle has higher capability of noise reduction compared to the one in Rayleigh distribution of speckle and the Kuan filter.
出处
《物理学报》
SCIE
EI
CAS
CSCD
北大核心
2007年第8期4565-4570,共6页
Acta Physica Sinica
基金
国家重点基础研究发展规划(批准号:2001CB309405)
国家自然科学基金(批准号:60574033)资助的课题.~~
