Learning Active Appearance Models from Image Sequences

Saragih, J. and Goecke, R.

One of the major drawbacks of the Active Appearance Model (AAM) is that it requires a training set of pseudo-dense correspondences. Most methods for automatic correspondence nding involve a groupwise model building process which optimises over all images in the training sequence simultaneously. In this work, we pose the problem of correspondence nding as an adaptive template tracking process. We investigate the utility of this approach on an audio-visual (AV) speech database and show that it can give reasonable results.

Cite as: Saragih, J. and Goecke, R. (2006). Learning Active Appearance Models from Image Sequences. In Proc. HCSNet Workshop on the Use of Vision in Human-Computer Interaction, (VisHCI 2006), Canberra, Australia. CRPIT, 56. Goecke, R., Robles-Kelly, A. and Caelli, T., Eds. ACS. 51-60.

(from crpit.com) (local if available)