遥感分类图像条带噪声的去除

Destriping Multispectral Image of Post Classification

在介绍了几种常用于TM、MSS、SPOT等多传感器遥感图像中条带噪声去除方法的基础上,提出了一种综合利用IDL语言和常用遥感软件(主要为ENVI、ERDAS等)对分类后图像进行条带处理的新方法。并以2005年北京市SPOT图像为试验数据,对该方法进行了尝试。结果表明,利用该方法可弥补一些条带噪声去除方法的弊端,有效地去除分类后图像上的条带噪音;同时避免了分类前期条带去除过程中对条带像元值的不正确计算,以及对图像上正确像元的影响而导致的后期遥感分类过程中的错分误分问题,从而可以有效地提高遥感分类精度。这种方法在其它多传感器遥感图像的条带噪声去除中也有很强的适用性。

A This paper discusses the methods previously used in striping removal of TM, MSS, SPOT and presents a new method aiming at post classification images using IDL, ENVI and ERDAS. And I choose 2005 SPOT image of Beijing as sample area to testify this method. The application results show that the new method can achieve a better result than the previously used methods mentioned in this paper in removing striping noise. The most important point is that this method can avoid some problems caused by striping removal process before supervised classification. They include calculating the striping pixels value incorrectly and the effects on non-striping pixels. So it can improve the precision of remote sensing classification. The new method is also applicable in striping removal of other muhisensor remote sensing data.