Active Appearance Model

from Wikipedia, the free encyclopedia

Active Appearance Model (AAM) is an image processing method developed by Cootes, Edwards and Taylor , with which classes of deformable objects such as faces or medical images of organs can be characterized particularly well.

The Active Appearance Model is a further development of the active contours ( Snakes ) and the Active Shape Model .

Procedure

The active appearance concept consists of three parts:

  1. A shape model, in which distinctive points (landmarks) of the object image are brought into a uniform (medium or typical) shape.
  2. In a second step, the original image is brought into this uniform form using image warping . The shape of all object images is thus standardized (“shape-free model”) and the variations in the images are due to differences in appearance, which can now be modeled separately. The core of the modeling in the first two steps is the principal component analysis , with which the typical modes of variation within an object class can be learned from a set of training images.
  3. Finally, the term “Active” means that an algorithm has been developed with which one can automatically search for the best adaptation of the AAM in a new, unknown image. The AAM can transform itself according to the learned modes of form and appearance.

literature

  • TF Cootes, GJ Edwards, CJ Taylor: “Active Appearance Models”, in: H. Burkhardt, B. Neumann (Eds.), Proceedings of the European Conference on Computer Vision , Vol. 2, pp. 484-498, Springer 1998
  • TF Cootes, GJ Edwards, CJ Taylor: "Active Appearance Models", in: IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 23, no. 6, June 2001, pp. 681-685

Web links