SAR图像多视处理中相邻子带重叠宽度的最优化

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

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。

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 muhilook processing, the improved method has a better performance in speckle suppression.