基于2-D熵的机载SAR图像分割迭代算法研究

The Recurring Algorithms Research of Airborne SAR Images Segmentation Based on Two-Dimensional Entropy

机载合成孔径雷达(SAR)图像可用于检测识别机场跑道。由于SAR图像包含大量噪声,算法抗噪性能对于SAR图像处理很重要。2-D阈值方法的分割效果比传统的一维方法要好,特别是对受噪声影响的图像。然而,2-D阈值方法耗时长,难以实用。因此,本文在现有的2-D熵迭代分割算法的基础上提出了一种基于二维直方图统计特点的最大熵阈值分割算法。通过大量试验证明,本文提出的方法不仅获得了理想的分割效果,而且进一步提高了二维阈值选取速度。

Airborne synthetic aperture radar（SAR） image could be used to detect runways. However, it is noised. So noise immunity of algorithms is important for SAR image processing. Two - dimen- sional（2 -D）thresholding methods can get better segmenting result than the traditional one -dimensional（ 1 - D）methods, especially in noised images. However, 2 - D thresholding methods need too much computional time to satisfy practical applications. To speed up the 2 - D entropy thresholding, a new method based on the statistical characteristic of 2 - D histogram was proposed based on the existing recurring algorithm of 2 - D entropy thresholding. The experiments showed that the proposed method could not only get the satisfied results, but also accelerate the 2 -D thresholding on 2 -D Renyi＇s entropy thresbolding method.