引言合成孔径雷达与其它侦察系统一样能提供极高的原始数据率。因此,对成象进行自动分析是人们希望的。然而,成象的相干性意味着数据受斑点噪声污染。这种噪声的标准偏差极高,而这正是任何一种图象分析算法都存在的问题。人们已经提出许多成功地应用于低噪声非相干数据环境的算法(参见光学),但应用于高噪声合成孔径雷达成象环境的算法尚很少见到。文献〔1〕提出了一种方法,该方法试图按照一个预定的图象模型来处理图象。只要该模型是精确的,这种方法就能很好地工作。但 Introduction Synthetic aperture radars offer the same high raw data rates as other reconnaissance systems. Therefore, it is hoped that the automatic analysis of the image will be done. However, the coherence of the image means that the data is contaminated with speckle noise. The standard deviation of this noise is extremely high, and this is exactly the problem with any image analysis algorithm. Many algorithms that have been successfully applied to low-noise, non-coherent data environments have been proposed (see Optics), but algorithms for high-noise synthetic aperture radar imaging environments are seldom seen. Document [1] proposes a method which attempts to process an image according to a predetermined image model. As long as the model is accurate, this approach works well. but