Summary: Retinal Vessel Extraction with the Image Ray
Alastair H. Cummings and Mark S. Nixon
University of Southampton, Southampton, SO17 1BJ, UK
Abstract. Extraction of blood vessels within the retina is an important
task that can help in detecting a number of diseases, including diabetic
retinopathy. Current techniques achieve good, but not perfect perfor-
mance and this suggests that improved preprocessing may be needed.
The image ray transform is a method to highlight tubular features (such
as blood vessels) based upon an analogy to light rays. The transform has
been employed to enhance retinal images from the DRIVE database, and
a simple classification technique has been used to show the potential of
the transform as a preprocessor for other supervised learning techniques.
Results also suggest potential for using the ray transform to detect other
features in the fundus images, such as the fovea and optic disc.
Automated detection of blood vessels within eye fundus images is an important
step in aiding diagnosis of a number of diseases, diabetic retinopathy in par-
ticular. The possible application of vision techniques to this problem was first