Abstract

According to the diffraction tomography data, the efficiency of minimization algorithms used to reconstruct the displacement field of a defect in a crystal depends on the presence of a noise component and implies preliminary use of noise filtering algorithms. The quality of projection image filtering has been estimated using the root-mean-square deviations of the intensity of a denoised image from the intensity of the initial noiseless image in a fixed rectangular neighborhood of the defect under study and beyond it. A comparison of these quantities, calculated after application of different noise filtering algorithms, has shown that their minimum values are obtained simultaneously using a guided image filter. The 3D reconstruction based on the projection images denoised in this way has significantly improved the quality of reconstructing the defect displacement field as compared with the results based on noisy unfiltered images.

中文翻译：

---

降低统计噪声水平和衍射层析成像二维图像滤波的理论

摘要

根据衍射断层扫描数据，用于重建晶体缺陷位移场的最小化算法的效率取决于噪声分量的存在，并暗示了噪声滤波算法的初步使用。已经使用被研究缺陷的固定矩形邻域中以及超出其范围的降噪图像的强度与初始无噪声图像的强度的均方根偏差来估计投影图像滤波的质量。在应用了不同的噪声过滤算法后，对这些数量进行了比较，结果表明，它们的最小值是使用引导图像滤波器同时获得的。