SAR图像MAP降噪的精细研究

Refined Research on MAP Filtering for SAR Images

本文推导出基于最大后验概率(MAP)滤波的一般形式,给出不同噪声分布和真实图像先验分布条件下的MAP滤波方程.从滤波方程在特定区间上解的分布情况以及区域统计特性分类两方面分析了MAP降噪性能,由此给出了MAP滤波的阈值表达形式.最后给出合成孔径雷达(SAR)图像的MAP降噪试验以及噪声滤除能力的量化指标.为了消除噪声强度对试验结果的影响,全面反映MAP降噪性能,本文给出了降噪能力随噪声大小的动态变化关系.结果表明,真实图像的先验分布对MAP滤波性能有着直接的影响,不合理的先验分布假定会严重降低MAP滤波的降噪能力.

General form of maximum a posteriori （MAP） filter is deduced, and some MAP equations are obtained by using different speekle distribution and prior distribution. Based on the distributing status of real roots in the interval and the statistics classification of synthetic aperture radar （SAR） images, threshold expressions of MAP filters are presented. Lastly de-speckling experiments on real SAR image and their quantitative measures are provided. In order to eliminate the influence of noise intensity to experimental results and eompletely reflect the performanee of MAP filters, dynamic relation of the de-speckling capability to noise varianee is suggested. Results show that MAP filtering performance is directly determined by the prior distribution. When prior distribution is chosen improperly, MAP filtering performance is degraded seriously.