MARS - Models of Appearance, Registration and Segmentation - Segmentation in Relation to Models and NRR

The following observations motivate work towards the inclusion of segmentation in the process of automatic (appearance) model-building:

Registration aims at getting segments to overlap
Maybe segment one image and propagate to all others via model
Use segmentation to improve evaluation (similar to label)
Conversely, improve segmentation using dynamics of models (built using NRR)
Look at many images (sequence) of modelled images, project back from model to get segmentation
The approach gives registration, model, and segmentation directly from the raw data
Segmentation is arbitrary (not rational but data-driven), just like choice of landmark points
Create small compartmentalised models of constituent segments
When model is built, see variances in mean and optimise points, then re-run model-building
This approach can progressively improve identification of landmarks
Hence, better models can be built automatically
Expensive approach as models are built over and over again to get refined

[From the May 31st 2005 Progress Report]