PDF version of this entire document

Sensitivity

Figure 11 compares the sensitivities of the different methods of evaluating NRR. This shows that specificity is more sensitive (is able to detect smaller misregistrations) than either generalisation or the overlap-based approach. Note from the error bars that these differences are statistically significant. Maximum sensitivity is achieved with shuffle radii of $1.5$ and $2.1$. Generalisation is shown to be a valid but not particularly sensitive measure of misregistration.

Figure 12: Appearance model which was built automatically by group-wise registration. First mode is shown, $\pm 2.5$ standard deviations.