||Many eye diseases cause noticeable loss of vision only at an advanced
stage, when curing is difficult or no longer possible.
The main objective of our research project is to develop automatic image
analysis methods to classify glaucoma and other eye diseases in order to
make early diagnosis and treatment possible.
Input images are acquired by different modalities (such as color fundus
photographs or topography images from confocal scanning laser tomography)
and show different features that should be combined for classification.
First, multimodal input images have to be registered, corrected for
illumination artifacts and the intensity values should be normalized.
Then, regions of interest (e.g., the optic nerve head and blood vessels)
can be segmented for direct feature measurements (e.g., sizes, distances,
shape descriptors) or for appearance-based analysis.
Finally, the extracted features and prior expert-knowledge can be used
for classification and computation of risk factors.