Base-line Models

$next$ $up$ $previous$ $contents$ $index$
Next: Similarity Measures Up: Initial Exploration Previous: Analysis of Warps Contents Index

Base-line Models

As a starting point for model construction and understanding, 10 regular data instances, which have been used in many of the experiments in up-coming sections, were generated and their statistical models built.

**Figure:** Shape model of 10 data instances at the start. The four principal modes are shown with up to standard deviations away from the mean.
$\includegraphics[%% scale=0.65]{./Graphics/shape_model_10.eps}$

It can be seen in Figure $[*]$ that shape is rather stable. This is because under these specific experiments, it was mainly (if not only) intensity that was used to create an appearance model^7.9 (more on this on page $[*]$ in this very same chapter).

**Figure:** Intensity model of 10 data instances at the start. The two principal modes are shown with up to standard deviations away from the mean.

Intensity models show some of the effects of height being changed, width varying and bump position moving from left to right. However, it is all rather fuzzy and the variation modes combine in a mysterious way. This is in fact why real correspondences need to be learned.

**Figure:** Combined (shape and intensity) model of 10 data instances at the start. The two principal modes are shown with up to standard deviations away from the mean.

The figure above shows the combination of intensity and shape. It is not yet too clear how to analyse it, but it resembles the intensity model which is the greater component that the combined model accounts for. Shape was quite static so it is expected to be merely invisible in the figure above.

$next$ $up$ $previous$ $contents$ $index$
Next: Similarity Measures Up: Initial Exploration Previous: Analysis of Warps Contents Index

2004-08-02