2D SEM (White) Grain Segmentation

This receipt expects a 2D image acquired from SEM. The image should have lots of small white grains that may be look a bit connected together, and yet they should be separate. Therefore the white grains cannot be easily segmented by the traditional thresholding method, as multiple grains would be incorrectly identified as one grain. Therefore a more accurate approach would require a more advanced image processing to identify each individual grain.

The recipe first upscales the image, and then identify the white grains using traditional thresholding method. Then after the recipe locates the vague separation area among the grains by using tophat method, the separation areas area subtracted from the previously identified white grains. Labeling and closing is used to make sure the grains are completely filled. Then the result are further separated according to its morphology before the final label analysis is done to calculate their physical attributes such as volume and surface area.