PDF version of this document

Future Aims and Objectives

The objectives remain similar to the original ones, yet new additional problems must be tackled first. The following are some of the more important challenges:

1. Improving the search method.
2. Including a component for model discrepancy in the objective function.
3. Obtaining plausible registration results while constructing appearance models that encapsulate the expected variation.
4. Learning about the robustness of the method.
5. Investigating the quality of the results.
6. Re-writing the objective function to calculate a proper MDL term².
7. Extension of the method to 2-D and 3-D.
8. Conduct experiments to show the advantages of:
   • Registration based on an MDL criterion.
   • Automatic construction of appearance models³.