Analysis of the Algebraic Tails of GF-3 SAR Images

Abstract

GF-3 is the synthetic aperture radar (SAR) satellite with the largest number of imaging modes in the world, which operates in 12 imaging modes. Owing to the 1-m resolution imaging mode, GF-3 satellite has become the highest resolution satellite system in the world for C-band multipolar SAR satellites. The high-resolution SAR images taken by the GF-3 satellite are rich in information. However, due to the coherent imaging system, GF-3 SAR images are affected by speckle, which degrades the quality of images seriously and makes the postprocessing of images such as edge detection, image segmentation, and target recognition extremely difficult. In this paper, we analyzed the characteristics of GF-3 SAR images from two aspects of samples and distributions. It can be seen that the GF-3 SAR images have algebraic tails. Furthermore, the above characteristics of GF-3 SAR images are verified by related experiments including the heavy-tailed Rayleigh distribution-based modeling and the maximum a posteriori (MAP) filter using the heavy-tailed prior distribution. Compared with the traditional models, GF-3 SAR images are well modeled by the heavy-tailed Rayleigh distribution that owns clear characteristics of sharp peak and heavy tail. This means that the GF-3 SAR images are statistically sharp-peaked and heavy-tailed. Additionally, the MAP filter with the heavy-tailed prior distribution achieves better performance in comparison with the traditional filters such as Lee filter and Gamma MAP filter, which is attributed to the choice of heavy-tailed prior distribution with the sharp peak and heavy tail. Therefore, based on the experiments above, we verify that the GF-3 SAR images own sharp-peaked and heavy-tailed statistical characteristics.