This project was realised in cooperation with Departement of Biomedical Engineering, Brno University of Technology. Aim of this project is to develope system consisting from optical microscope and appropriate software in order to automatically detect and classify different types of leucocytes. Result of this work is automated generation of blood picture in much more cheaper way then today's blood-picture machines.
You can see my another computer vision projects:
Overlapping object segmentation and classification
Self-Locating by Pattern Tracking
My work is particular part of the whole project, which provides detection of leucocyte's outline and then realizes its classification to classes based on their characteristical features.
First part (outline detection) is programmed in very untraditional way. We uses Canny's filter disadvantage - its noise sensitivity - as advantage in order to distinguish smooth areas (erytrocytes and background) from rough noisy areas (leucocytes).
In second part, we computes particular features of each leucocyte and then classify them into the classes by decision-tree.
Here you can find some documents about solving out this task. Only presentation slides are available.
size: 871.79kB, last modified: 19.08.2014