Vessel Segmentation and Tracking Using a Two-Dimensional Model

The segmentation and analysis of blood vessels in retinal images is of immense interest for study of diseases involving vasculature changes. An algorithm to segment blood vessels in colour retinal images by tracking of vessels is described. This algorithm proceeds by fitting a physically inspired two-dimensional model of the vessel profile to a local region of the vessel. By fitting in this manner a number of parameters, such as diameter and orientation, of the local vessel segment, can be accurately measured as tracking proceeds. A modification to the model that enables tracking of tortuous vessels is described. A method to detect vessel branches is also described. Illustrations of the vessel tracking working on retinal images are given.