REGISTRATION EXPERIMENTS

The format of the notes has changed and the writings are now more concise and focused. All the individual files, settings and results are available from the experiments archive.

width


07122004-1: Long optimisation with 10 images, group-wise.

07122004-2: Long optimisation with 10 images, pair-wise.

Intent: Comparing performance of pair-wise registration and group-wise registration based on a long optimisation.

Conclusion: Group-wise registration sometimes does better than pair-wise registration, but not always. Conclusion:

width


09122004-1: Group-wise registration of 10 images, all at resolution level 2

09122004-2: Group-wise registration of 10 images, all at resolution level 1

Intent: Checking the resolution effects in large group-wise registration.

Conclusion: still running.

width


10122004-1: MI, Rigid and affine registration of 10 images, resolution 64x64.

10122004-2: MI, Rigid and affine registration of 10 images, resolution 128x128.

10122004-3: MI, Rigid and affine registration of 10 images, resolution 256x256.

*At this stage, it was important to see what happens below 64x64

width


10122004-4: MI, Rigid and affine registration of 10 images, resolution 32x32.

Intent: Investigating MI and NMI (in 3-D) as resolution of images is changed.

Conclusion: Subjective. See results.

width


11122004-4: MI, Rigid and affine registration of 10 images, resolution 32x32.

Intent: Seeing if MI and NMI are stable at the solution Description: 2 images that are the same are rigidly and affinely aligned using MI and (later on) NMI.

Conclusion: MI and NMI are stable at the solution.