论文部分内容阅读
合成孔径雷达(SAR)图像的低信噪比和乘性相干斑噪声给SAR图像的边缘检测带来了极大的困难。通过引入广义高斯(GG)分布作为局部均值功率的先验分布模型,给出了局部均值功率在最大后验概率(MAP)意义下的最优估计,进而提出一种新的SAR图像边缘比率检测算子。利用以梅林变换为基础的对数累积量(MoLC)方法估计GG分布的参数,在此基础上给出一种局部均值功率MAP估计和GG分布参数估计的联合迭代求解方法。利用SAR实测数据对本文提出的边缘检测算子进行仿真验证,并将其与平均比率(RoA)算子和指数加权均值比(ROEWA)算子进行了对比,结果表明该算子可以有效克服相干斑噪声的影响,边缘定位准确且虚假边缘明显减少。
The low signal-to-noise ratio and multiplicative speckle noises of Synthetic Aperture Radar (SAR) images bring great difficulties to the edge detection of SAR images. By introducing a generalized Gaussian (GG) distribution as a priori distribution model of local mean power, an optimal estimation of local mean power under the maximum a posteriori probability (MAP) is given. Then a new SAR edge detection ratio operator. The parameter of GG distribution is estimated by using the method of logarithmic cumulant (MoLC) based on Merlin transform. Based on this, a joint iterative method for local MAP estimation and GG distribution estimation is given. The edge detection operator proposed in this paper is validated by the measured SAR data and compared with the RoA operator and the ROEWA operator. The results show that the operator can effectively overcome the coherence The effect of speckle noise, accurate edge location and false edges significantly reduced.