Glaucoma is one of the leading causes for blindness as it induces a degeneration of retinal nerve fibers. The occurence of the disease is also accompanied by changes of the optic nerve head's (ONH) appearance. Based on a quantitative characterisation of the ONH several glaucoma indices were developed. However, most of these techniques rely on geometric measurements or parameters of the ONH such as diameter or area. From our point of view, the established techniques only utilize a sparse characterization of the ONH. Thus, we propose to describe the inter subject variability of the ONH by dense deformation fields produced by non-rigid registration. It represents the first step towards high resolved statistically modelling of ONH variability.

Rigid registration is used to eliminate translational variations between the ONH. It is applied on gradient magnitude fundus images. This representation emphasizes the ONH as region of interest and suppresses the individual vessel tree in case of large filter support and is less sensitive to luminosity variations .

In order to capture the inherent ONH variability, we apply an adapted non-rigid registration approach. It is extended by a regularisation term approving radial deformations to ensure a physiological, circular mapping of the corresponding ONH sectors. Since the vessel tree is highly individual like a fingerprint, it is masked out during the calculation of the distance measures.