A good computer vision algorithm can't be complete without great, robust capabilities as well as wide generalization and a solid math foundation. All these features accompany the work mainly developed by Tim Cootes with Active Appearance Models. This chapter will teach you how to create an Active Appearance Model of your own using OpenCV as well as how to use it to search for the closest position your model is located at in a given frame. Besides, you will learn how to use the POSIT algorithm and how to fit your 3D model in the "posed" image. With all these tools, you will be able to track a 3D model in a video, in real time; ain't it great? Although the examples focus on head pose, virtually any deformable model could use the same approach.
As you read the sections, you will come across the following topics:
Active Appearance Models overview
Active Shape Models overview
Model instantiation—playing with the Active Appearance Model