PDF version of this entire document
Non-rigid registration (NRR) of a group of images [6,4] establishes correspondence among the set. That correspondence can also be assessed using measures of overlap between labels in the images [2] or by creating degraded test sets [5] and attempting to recover a correct solution. A new method, which has already been validated in 2-D, is proposed as the means for assessing 3-D registration of images of the brain. Ultimately, different NRR algorithms can be compared and ranked based on their performance.