Experimental Tests

Three separate tests of the similarity functions and registrations will be performed:

1. Each shape having a constant intensity.
2. Shape 2 having a linear intensity gradient (representing variation of the object's density).
3. As for test 2 but with bias field (extra linear intensity gradients) added to the measured image, $Y$. The amount of bias field was chosen randomly for each trial from a uniform distribution such that the intensity variation did not exceed $\pm$10% of the mean intensity of shape 2.

In each case a range of SNR values is tried, each using a set of 100 instances of measurement noise.

Furthermore, the field of view of the measurement, $Y$, was shifted along the model object, as illustrated in figure . This allowed the sensitivity of the similarity functions with respect to changing fields of view to be measured.