Abstract The backscattered electron (BSE) mode in the scanning electron microscope has been frequently used to characterize phases with different chemical compositions in metals and alloys. For phase quantification, the ASTM E562 standard describes the manual point counting method, which is rather inefficient. New methods use automated threshold-based digital image segmentation in which a threshold grey value is defined within a greyscale range in order to subdivide the phases into black and white. However, these methods are either computationally expensive and parameter-dependent or limited to images that have a bimodal greyscale distribution and an even background grey intensity. This is frequently not the case for BSE images. In this work, the rolling ball algorithm is used to correct background for the first time in BSE images. An approach is proposed to determine a bias-free threshold grey value aided by selected electron backscatter diffraction (EBSD) studies. The proposed methodology is implemented in MATLAB and used to quantify the α phase fractions in a β-metastable titanium alloy. The results show the capability of the method in correcting the background in BSE images and quantifying phase fractions at comparatively low labor and computational costs. Further, the proposed procedure may have a potential significance for machine learning image analysis algorithms.

中文翻译：

---

滚球算法在背散射电子图像相体积分数测量中的应用

摘要 扫描电子显微镜中的背散射电子 (BSE) 模式经常用于表征金属和合金中具有不同化学成分的相。对于相位量化，ASTM E562 标准描述了手动点计数方法，该方法效率相当低。新方法使用基于阈值的自动数字图像分割，其中在灰度范围内定义阈值灰度值，以便将相位细分为黑色和白色。然而，这些方法要么计算成本高且依赖于参数，要么仅限于具有双峰灰度分布和均匀背景灰度强度的图像。BSE 图像通常不是这种情况。在这项工作中，滚球算法首次用于 BSE 图像中的背景校正。提出了一种通过选定的电子背散射衍射 (EBSD) 研究来确定无偏置阈值灰度值的方法。所提出的方法在 MATLAB 中实施，用于量化 β 亚稳态钛合金中的 α 相分数。结果显示了该方法在校正 BSE 图像中的背景和以相对较低的劳动力和计算成本量化相位分数方面的能力。此外，所提出的程序可能对机器学习图像分析算法具有潜在意义。结果显示了该方法在校正 BSE 图像中的背景和以相对较低的劳动力和计算成本量化相位分数方面的能力。此外，所提出的程序可能对机器学习图像分析算法具有潜在意义。结果显示了该方法在校正 BSE 图像中的背景和以相对较低的劳动力和计算成本量化相位分数方面的能力。此外，所提出的程序可能对机器学习图像分析算法具有潜在意义。