Automated grain sizing using mark-based watershed algorithm
Han-Sheng Chuang, Chao‐Hung Lin · 2012
This paper presents a method based on mark-based watershed algorithm to automatically extract grains and determine grain sizes from images. Markers generated by the proposed approaches represent rough locations of grains and aperture. Instead of selecting pixels with local minima as marks, we select markers with a priori knowledge, enabling our approach to efficiently ease the problem of over-segmentation in traditional watershed algorithm. The grains patches are extracted correctly after merging the other fragmental patches using both the color and shape properties. Finally, the distribution of grain size is calculated by fitting an ellipse for each detected grain.