Active Appearance Models overview

In few words, Active Appearance Models are a nice model parameterization of combined texture and shape coupled to an efficient search algorithm that can tell exactly where and how a model is located in a picture frame. In order to do that, we will start with the Active Shape Models section and will see that they are more closely related to landmark positions. A principal component analysis and some hands-on experience will be better described in the following sections. Then, we will be able to get some help from OpenCV's Delaunay functions and learn some triangulation. From that we will evolve to applying piecewise affine warps in the triangle texture warping section, where we can get information from an object's texture.

As we get enough background to build a good model, we can play with the techniques in the model instantiation section. We will then be able to solve the inverse problem through AAM search and fitting. These by themselves are already very...