SAR图像多视处理中相邻子带重叠宽度的最优化

多视处理是合成孔径雷达（SAR）图像纹斑噪声抑制的一种基本方法。在这种方法里，用一组多视滤波器将图像的多普勒谱分成若干个子带，相邻子带之间可以有一定的重叠；首先由每个子带生成一幅子图像，然后再将这些子图像非相干平均起来。在传统的多视处理中，相邻子带重叠宽度的选择是经验性的。提出了相邻子带重叠宽度的一种最优化算法。采用图像的标准差与均值的比值作为衡量纹斑噪声强弱的一种测度，选择相邻子带重叠宽度以最小化这一测度。实验证明，在分辨率保持不变的条件下，该算法在纹斑噪声抑制方面优于传统算法。

Optimization of overlap width between adjacent sub-bands in multilook processing of SAR images

Multilook processing is a basic method for speckle suppression of synthetic aperture radar(SAR) images.In this method,the Doppler spectrum is divided into multiple sub-bands,and adjacent sub-bands can overlap each other.Each sub-band is used to generate a look,and all the looks are averaged incoherently to obtain the final image.In the traditional multilook processing,the overlap width between adjacent sub-bands is chosen empirically.An optimization algorithm of the overlap width between adjacent sub-bands is introduced in this paper.This paper defined a measure of speckle intensity and choose the overlap width between adjacent sub-bands to optimize this measure.Experiments show that under the same resolution,compared to the traditional multilook processing,the improved method has a better performance in speckle suppression.