Counting red blood cells in blood smear.
The goal of this project is to count the number of blood cells in a 2D image of a blood smear.
I use several steps of image pre-processing to obtain a binary image from which it is easy to count the cells by counting the white zones in the image.
The most accurate method I found is a pattern matching approach.
-
I take a reference cell, and compute its convolution against the complete image. Higher intensity zones corresponds to cells center.
-
Then, I apply a threshold at a given percentage of the histogram to get a binary image where the cells are white circles and the rest is black.
-
Finally, I count the white zones in the binary image.
This project was part of my Master Of Science curriculum at Centrale Nantes.
This project is no longer updated.