Name
Abstract | ||
|---|---|---|
Active appearance models (AAMs) are widely used to fit statis tical models of shape and appearance to images, and have applications in segmentation, tracking, and classification of structures. A limitation of AAMs is that they are not robust to a large set of gross outliers. Using a robust kernel can help, but there are potential problems in determining the correct kernel scaling pa- rameters. We describe a method of learning two sets of scaling parameters during AAM training: a coarse and a fine scale set. Our algorit hm initially applies the coarse scale and then uses a form of deterministic annealing to reduce to the fine outlier rejection scaling as the AAM conver ges. The algo- rithm was assessed on two large datasets consisting of a set of faces, and a medical dataset of images of the spine. A significant improve ment in accu- racy and robustness was observed in cases which were difficul t for a standard AAM. |
Year | Venue | Keywords |
|---|---|---|
2007 | BMVC | active appearance model |
Field | DocType | Citations |
Kernel (linear algebra),Computer vision,Pattern recognition,Segmentation,Computer science,Outlier,Active appearance model,Robustness (computer science),Deterministic annealing,Artificial intelligence,Scaling | Conference | 3 |
PageRank | References | Authors |
0.91 | 9 | 3 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Martin G. Roberts | 1 | 27 | 2.36 |
| Timothy F. Cootes | 2 | 4358 | 579.15 |
| Judith E. Adams | 3 | 53 | 5.57 |