ABSTRACT
Active Appearance Models
Tim Cootes, University of Manchester
An Appearance Model contains statistical models of both the shape and the
texture of a structure in an image. They can be generated from training
sets of images labelled with the positions of landmark points around
the structures of interests. They can be used to reconstruct synthetic
images of the structures. By varying the parameters between limits
learnt from the training set new examples, similar to those in the
training set, can be generated. The Active Appearance Model algorithm
allows us to match a model to a new image rapidly. The method works
by using a vector of differences between the current model appearance
and the target image appearance to drive the model parameters toward
a better match.
In this talk I will describe the models and search algorithm, giving examples
of its application to images of faces and medical images.
Computer Science Home Page
Maintained by rbennett@cs.ucl.ac.uk