一种面向高异质性极化SAR图像的等效视数非监督估计方法

Unsupervised Estimation of the Equivalent Number of Looks in PolSAR Image with High Heterogeneity

等效视数(ENL)是极化SAR多视数据统计模型的重要参数。而一些极化SAR图像的自动化应用中,需要在没有人工干预下实现ENL非监督估计。现有的等效视数非监督估计方法在异质程度较高的图像中就难以得到准确估计结果。针对这一问题,该文提出一种将混合区域剔除与纹理信息聚类相结合的等效视数非监督估计方法,有效地减弱了地物混合及纹理两类主要异质因素对估计结果的影响。通过仿真数据和不同复杂度的实际图像验证了该方法的有效性。

Equivalent Number of Looks （ENL） is an important parameter in statistical modeIiing of muiti-iook Polarimetric SAR （PolSAR） data. In some automated applications of PolSAR images, it is necessary to estimate the ENL in an unsupervised way without any manual intervention. The existing unsupervised estimation of ENL can not obtain accurate estimates for the images with high heterogeneity. To address this issue, a novel unsupervised estimation method is proposed here. It combines the mixture elimination and clustering based on texture, which reduces the effect of two main heterogeneity factors, mixture and texture. The validity of this method is evaluated with simulated and real data of different complexity.