Assessing Non-Rigid Registration
Without Using Ground Truth
R. S. Schestowitz, V. Petrovic, C. J. Twining,
T. F. Cootes, and C. J. Taylor
We present a method for assessing the performance of non-rigid
registration algorithms, without the need for any form of ground truth.
The method exploits the fact that, given a set of non-rigidly registered
images, a generative statistical appearance model can be constructed.
The quality of the model depends on the quality of the registration,
and can be evaluated by comparing images sampled from it with the
original image set. We derive indices of model specificity and generalisation,
and show that they demonstrate the loss of registration as a set of
correctly registered images is progressively perturbed. We also show
an application of the method to comparing the performance of three
different registration algorithms.