A method of segmenting leukocytes on images of blood smears

摘要

This article explains the features of a method for segmenting microscopic images of blood specimens, in which the main types of leukocytes are discriminated and counted. The segmentation is based on threshold classification of the image of a blood smear according to brightness, hue, and saturation, using morphological processing of binary masks, calculation of the objects' parameters such as fill factor, eccentricity, and equivalent diameter, and calculation of the objects' texture characteristics. The method is stable against the presence of noise on the image, as well as against small changes of the color characteristics of the cells.