In this paper, an image processing technique is introduced to measure the grain size and their distributions from the SEM image of copper oxide (CuO) and titanium dioxide (TiO₂) doped sintered alumina ceramics accurately. The noise present in SEM image is removed by applying low pass Gaussian filter followed by suppression of regional minima over a threshold. The clarity of individual grains and grain boundaries have been done by applying Watershed transform to this preprocessed SEM image. Morphological operations like dilation and erosion are used to make the grain‐boundary edges clear and continuous. The individual grain size in µm scale is measured from the pixel length of the rectangular bounding box drawn around the segmented grain. The normal Gaussian type distribution of grain size is observed in both CuO‐ and TiO₂‐doped grains in SEM image. The average grain size of CuO‐doped alumina grains (2.24 µm) is very close to G₅₀ value (2.17 µm), but G₅₀ value of TiO₂‐doped grains (8.59 µm) is slightly higher than its average grain size (7.96 µm). The proposed algorithm is compared with linear intercept method and the grain sizes obtained are very close to each other.

中文翻译：

---

通过图像处理技术评估烧结陶瓷材料微观结构的新方法

本文介绍了一种图像处理技术，用于从氧化铜（CuO）和二氧化钛（TiO ₂）准确地掺杂烧结氧化铝陶瓷。通过应用低通高斯滤波器，然后抑制超过阈值的区域最小值，可以消除SEM图像中存在的噪声。通过将分水岭变换应用于此预处理的SEM图像，可以实现单个晶粒和晶界的清晰度。诸如膨胀和侵蚀之类的形态学操作用于使晶界边缘清晰连续。以微米为单位的单个晶粒尺寸是从围绕分段晶粒绘制的矩形边界框的像素长度测量的。在SEM图像中，在掺杂CuO和TiO _2的晶粒中均观察到了正态的高斯型晶粒尺寸分布。掺杂CuO的氧化铝晶粒的平均晶粒尺寸（2.24 µm）非常接近G ₅₀值（2.17 µm），但是G掺杂TiO _2的晶粒的₅₀值（8.59 µm）略高于其平均晶粒尺寸（7.96 µm）。将该算法与线性截距法进行比较，得到的晶粒尺寸非常接近。