合成孔径雷达图像的最小均方误差线性最优滤波

针对常用于合成孔径雷达(SAR)图像降噪的Lee滤波和Kuan滤波误差较大的问题,提出了基于最小均方误差(MMSE)准则的线性最优滤波.线性最优滤波通过把斑点噪声的乘性模型同时展开为一阶和二阶泰勒级数,然后使用MMSE准则获得线性滤波的统一模型,最后再对该统一模型使用MMSE准则而获得.线性最优滤波在所有的线性滤波中具有最低的滤波误差,因而具有最高的滤波精度.对某乡村和城区SAR图像的降噪实验表明:线性最优滤波对边缘细节的保留能力强于Kuan滤波,它对斑点噪声的滤除能力强于Lee滤波;与最大后验概率(MAP)滤波相比,线性最优滤波虽然具有较弱的边缘细节保留能力,但它对斑点噪声的滤除能力却强于MAP滤波.

Linear Optimal Filter with Minimum Mean Square Error for Synthetic Aperture Radar Images

A linear optimal filter is proposed based on the minimum mean square error(MMSE) criterion to solve the problem that the commonly used Lee and Kuan filters for synthetic aperture radar(SAR)images have bigger filtering errors.The multiplicative noise model of speckle is ex-panded into both the first-order and the second-order Taylor series at the same time.and then the MMSE criterion is used to deduce fl unified model of linear filters.The linear optimal filter is fi-nally obtained-by applying the MMSE criterion again to the unified model.The linear optimal fil-ter has the lowest filtering error and the highest filtering accuracy among all linear filters.The despeckling experiments on rural and urban SAR images show that the linear optimal filter has higher edge and fine detail preserving capacity than the Kuan filter,and has higher speckle sup-pression than the Lee filter.A comparison with the maximum a posteriori filter shows that the linear optimal filter has lower edge and fine detail preserving capacity,but has higher capability of speckle suppression.