Reconstruction From Aperture-Filtered Samples With Application to Scatterometer Image Reconstruction
Reconstruction From Aperture-Filtered Samples With Application to Scatterometer Image Reconstruction
复制标题
DOI：
10.1109/tgrs.2010.2086063
复制
发表时间：
2011-05
影响因子：
8.2
通讯作者：
B. Williams;D. Long
中科院分区：
文献类型：
--
作者：
B. Williams;D. Long
本文将散射计图像重建问题归结为离散噪声孔径滤波采样运算的逆运算。提出了孔径滤波采样方法，并与常规采样和非规则采样进行了对比。研究了无噪声孔径滤波样本的离散重建方法，并与传统的连续重建方法进行了比较。离散方法能够分析处理重建网格分辨率和采样和重建操作施加的有效分辨率。对噪声情况也进行了探讨。提出了一种基于最大后验概率(MAP)估计的重建估计器，用于从噪声散射计测量数据中恢复常规样本。该方法使得散射计噪声分布能够在重建操作中被适当地考虑。将MAP和常规重建方法应用于SeaWinds散射计和高级风散射计，分析了不同方法的有效分辨率。MAP方法产生的结果与公认的散射计图像重建(SIR)算法一致。MAP方法以增加噪声为代价显著提高了分辨率。虽然详细的噪声与分辨率权衡分析超出了本文的范围，但新的框架允许进行比SIR算法的特殊调谐参数更一般的处理。
This paper approaches scatterometer image reconstruction as the inversion of a discrete noisy aperture-filtered sampling operation. Aperture-filtered sampling is presented and contrasted with conventional and irregular sampling. Discrete reconstruction from noise-free aperture-filtered samples is investigated and contrasted with conventional continuous reconstruction approaches. The discrete approach enables analytical treatment of the reconstruction grid resolution and the effective resolution imposed by the sampling and reconstruction operations. The noisy case is also explored. A reconstruction estimator based on maximum a posteriori (MAP) estimation is proposed to recover the conventional samples from noisy scatterometer measurements. This approach enables the scatterometer noise distribution to be appropriately accounted for in the reconstruction operation. The MAP and conventional reconstruction approaches are applied to the SeaWinds scatterometer and the Advanced Wind Scatterometer, and the effective resolution of the different methods is analyzed. The MAP approach produces results consistent with the well-established scatterometer image reconstruction (SIR) algorithm. The MAP approach significantly enhances the resolution at the expense of increased noise. Although a detailed noise-versus-resolution tradeoff analysis is beyond the scope of this paper, the new framework allows for a more general treatment than the ad hoc tuning parameters of the SIR algorithm.