PDF version of this entire document

Validation using Warped Images

Given the 70 image sets described above, each with known average misregistration, 79#79, we investigated the relationship between 79#79 and Specificity, Generalisation, and Generalised Overlap, calculating the mean and standard error for each measure over the 10 warp instances for each value of 79#79.

For each misregistered image set, we calculated Specificity and Generalisation, as described in Section , using 34#34 = 15 modes of variation for the model and 82#82 = 1000 synthetic images drawn from a Gaussian distribution, as described in Section . This was repeated for values of shuffle radius, 66#66, of 1 (Euclidean distance), 1.5, 2.1 and 3.7, as defined in Section , corresponding to circular neighbourhoods contained within 1x1, 3x3, 5x5 and 7x7 pixel patches respectively. We also repeated these experiments with 2.5%, 5.0% and 10% Gaussian intensity noise added to the misregistered images, in order to investigate the sensitivity of the model-based measures to image noise.

Similarly, we calculated Generalised Overlap with volume, equal, inverse volume and complexity weightings, as defined in Section   calculating the mean and standard error for each measure over the 10 warp instances for each value of 79#79.