PDF version of this document

Similarity Measures

Figure shows one of the more useful measures for the model-based objective function. The MSD measure shows that the images become mutually similar as the model-based objective function proceeds, i.e. as the model of the data is minimised in its complexity. This confirms that a move is made in the right direction.

Figure: Mean MSD measures at each point during the model-based registration of 10 data instances.

Before investigating some of the different (and almost distinct) registration methods, it was worth looking at some measures and the way that these were affected as registration took place. Figure shows the measures of MSD and MI for a group of 2-D synthetic data^7.10. The large image shown in the figures is the reference image and the bottom plots show the measures for each of the 4 other images respectively.

Figure: 2-D Synthetic data generated and evaluated for similarity against the reference.