#### DMCA

## Active Appearance Models (1998)

### Cached

### Download Links

- [www.cse.msu.edu]
- [image.diku.dk]
- [www.ai.mit.edu]
- [www.cs.cmu.edu]
- [www.cs.cmu.edu]
- [www.cs.cmu.edu]
- [www.cs.cmu.edu]
- [www.wiau.man.ac.uk]
- [www.isbe.man.ac.uk]
- [www.isbe.man.ac.uk]
- [perso.telecom-paristech.fr]
- [perso.telecom-paristech.fr]
- [pages.cs.wisc.edu]
- [pages.cs.wisc.edu]
- [perso.telecom-paristech.fr]
- [www.isbe.man.ac.uk]
- [s10d.smb.man.ac.uk]
- [www.stat.ucla.edu]
- [www.wiau.man.ac.uk]
- DBLP

### Other Repositories/Bibliography

Venue: | IEEE Transactions on Pattern Analysis and Machine Intelligence |

Citations: | 2154 - 59 self |

### Citations

3880 | Eigenfaces for recognition - Turk, Pentland - 1991 |

1339 | Active Shape Models - Their training and application. - Cootes, Cooper, et al. - 1995 |

1108 | Visual learning and recognition of 3-D objects from appearance. - Murase, Nayar - 1995 |

695 | Active Contours. - Blake, Isard - 1998 |

563 | Efficient region tracking with parametric models of geometry and illumination.
- Hager, Belhumeur
- 1998
(Show Context)
Citation Context ...his allows us to find directions of rapid convergence even though the search space has very high dimensionality. The approach has similarities with ªdifference decompositionº tracking algorithms [7], =-=[8]-=-, [13], but rather than . T.F. Cootes and C.J. Taylor are with the Division of Imaging Science and Biomedical Engineering, University of Manchester, M13 9PT, UK. E-mail: {t.cootes, chris.taylor}@man.a... |

438 | Xm2vtsdb: The extended m2vts database,” in - Messer, Matas, et al. - 1999 |

237 | Representations of knowledge in complex systems. - Grenander, Miller - 1994 |

237 | Automatic Interpretation and Coding of Face Images using Flexible Models.
- Cootes, Lanitis, et al.
- 1997
(Show Context)
Citation Context ...x TermsÐAppearance models, deformable templates, model matching. 1 INTRODUCTION æ THE ªinterpretation through synthesisº approach has received considerable attention over the past few years [3], [6], =-=[11]-=-, [14]. The aim is to ªexplainº novel images by generating synthetic images that are as similar as possible, using a parameterized model of appearance. One motivation is to achieve robust segmentation... |

202 | A minimum description length approach to statistical shape modeling. - Davies, Twining, et al. - 2002 |

167 | Interpreting face images using active appearance models, in: - Edwards, Taylor, et al. - 1998 |

123 |
Reliable Head Tracking under Varying Illumination: An Approach Based on Registration of Texture Mapped 3D Models,º
- Cascia, Sclaroff, et al.
- 2000
(Show Context)
Citation Context ... a single template to deform under a variety of transformations (affine, projective, etc.). He chooses the parameters to minimize a sum of squares measure and essentially precomputes derivatives of the difference vector with respect to the parameters of the transformation. Hager and Belhumeur [8] describe a similar approach, but include robust kernels and models of illumination variation. Sclaroff and Isidoro [13] extend the approach to track objects which deform, modeling deformation using the low energy modes of a finite element model of the target. The approach has been used to track heads [10] using a rigid cylindrical model of the head. The Active Appearance Models described below are an extension of this approach [4], [1]. Rather than tracking a particular object, our models of appearance can match to any of a class of deformable objects (e.g., any face with any expression, rather than one persons face with a particular expression). To match rapidly, we precompute the derivatives of the residual of the match between the model and the target image and use them to compute the update steps in an iterative matching algorithm. In order to make the algorithm insensitive to changes in o... |

119 | reliable head tracking under varying illumination: An approach based on robust registration of texture-mapped 3d models - Fast |

107 | Face recognition using view-based and modular eigenspaces, - Moghaddam, Pentland - 1994 |

104 | Resynthesizing facial animation through 3d model-based tracking. - Pighin, Szeliski, et al. - 1999 |

96 | A multiview non-linear active shape model using kernel pca. - Romdhani, Gong, et al. - 1999 |

92 | Multidimensional morphable models : A framework for representing and matching object classes.
- Jones, Poggio
- 1998
(Show Context)
Citation Context ...meters that are useful for higher-level interpretation of the scene. Suitable methods of modeling photo-realistic appearance have been described previously, e.g., Edwards et al. [3], Jones and Poggio =-=[9]-=-, or Vetter [14]). When applied to images of complex and variable objects (e.g., faces), these models typically require a large number of parameters (50-100). In order to interpret novel images, an ef... |

90 | Statistical region snake-based segmentation adapted to different physical noise models. - Chesnaud, Refregier, et al. - 1999 |

67 | Multidimensional morphable models. - Jones, Poggio - 1998 |

65 | Learning to identify and track faces in image sequences. - Edwards, Taylor, et al. - 1998 |

47 | Generalized image matching: Statistical learning of physically-based deformations.
- Nastar, Moghaddam, et al.
- 1996
(Show Context)
Citation Context ...ject we match a model which can fit a whole class of objects. 2 BACKGROUND Several authors have described methods for matching deformable models of shape and appearance to novel images. Nastar et al. =-=[12]-=- describe a model of shape and intensity variations using a 3D deformable model of the intensity landscape. They use a closest point surface matching algorithm to perform the fitting, which tends to b... |

46 | ªFacial Analysis and Synthesis Using Image-Based Models,º
- Ezzat, Poggio
- 1997
(Show Context)
Citation Context ... Index TermsÐAppearance models, deformable templates, model matching. 1 INTRODUCTION æ THE ªinterpretation through synthesisº approach has received considerable attention over the past few years [3], =-=[6]-=-, [11], [14]. The aim is to ªexplainº novel images by generating synthetic images that are as similar as possible, using a parameterized model of appearance. One motivation is to achieve robust segmen... |

43 | Face recognition from unfamiliar views: Subspace methods and pose dependency. - Graham, Allinson - 1998 |

36 | Estimating coloured 3d face models from single images: An example-based approach - Vetter, Blanz - 1998 |

33 | der Malsburg. Tracking and learning graphs and pose on image sequences of faces - Maurer, van - 1996 |

30 | An algorithm for the learning of weights in discrimination functions using a priori constraints - Krüger - 1997 |

26 | From regular images to animated heads: A least squares approach. In - Fua, Miccio - 1998 |

23 | Learning novel views to a single face image. - Vetter - 1997 |

15 | Understanding pose discrimination in similarity space. - Sherrah, Gong, et al. - 1999 |

14 | Segmentation of cardiac mr images: An active appearance model approach. - Mitchell, Lelieveldt, et al. - 2000 |

14 | On utilising template and featurebased correspondence in multi-view appearance models - Romdhani, Psarrou, et al. - 2000 |

6 | Model-based initialisation for segmentation. - Hug, Brechbuhler, et al. - 2000 |

6 |
ªActive Appearance Models,º
- Cootes, Edwards, et al.
- 1998
(Show Context)
Citation Context ...nt model of the target. The approach has been used to track heads [10] using a rigid cylindrical model of the head. The Active Appearance Models described below are an extension of this approach [4], =-=[1]-=-. Rather than tracking a particular object, our models of appearance can match to any of a class of deformable objects (e.g., any face with any expression, rather than one persons face with a particul... |

6 |
ªActive Shape ModelsÐTheir Training and Application,º
- Cootes, Taylor, et al.
- 1995
(Show Context)
Citation Context ...uire face images marked with points defining the main features (Fig. 1). We apply Procrustes analysis to align the sets of points (each represented as a vector, x) and build a statistical shape model =-=[2]-=-. We then warp each training image so the points match those of the mean shape, obtaining a ªshape-free patchº (Fig. 1). This is raster scanned into a texture vector, g, which is normalized by applyin... |

5 |
ªFast, Reliable Head Tracking under Varying Illumination: An Approach Based on Registration of Texture-Mapped 3D Models,º
- Cascia, Sclaroff, et al.
- 2000
(Show Context)
Citation Context ... and Isidoro [13] extend the approach to track objects which deform, modeling deformation using the low energy modes of a finite element model of the target. The approach has been used to track heads =-=[10]-=- using a rigid cylindrical model of the head. The Active Appearance Models described below are an extension of this approach [4], [1]. Rather than tracking a particular object, our models of appearanc... |

4 |
ªProjective Registration with Difference Decomposition,º
- Gleicher
- 1997
(Show Context)
Citation Context ...ne. This allows us to find directions of rapid convergence even though the search space has very high dimensionality. The approach has similarities with ªdifference decompositionº tracking algorithms =-=[7]-=-, [8], [13], but rather than . T.F. Cootes and C.J. Taylor are with the Division of Imaging Science and Biomedical Engineering, University of Manchester, M13 9PT, UK. E-mail: {t.cootes, chris.taylor}@... |

3 |
ªActive Blobs,º
- Sclaroff, Isidoro
- 1998
(Show Context)
Citation Context ...llows us to find directions of rapid convergence even though the search space has very high dimensionality. The approach has similarities with ªdifference decompositionº tracking algorithms [7], [8], =-=[13]-=-, but rather than . T.F. Cootes and C.J. Taylor are with the Division of Imaging Science and Biomedical Engineering, University of Manchester, M13 9PT, UK. E-mail: {t.cootes, chris.taylor}@man.ac.uk. ... |

2 |
ªFace Recognition Using the Active Appearance Model,º Proc.
- Edwards, Taylor, et al.
- 1998
(Show Context)
Citation Context ...ers of branches, but can be used for various organs which exhibit shape variation but not a change in topology. We have tested the method extensively with images of faces and with medical images [1], =-=[5]-=- and have obtained encouraging results in other domains. Although an AAM search is slightly slower than Active Shape Model search [2], the procedure tends to be be more robust than ASM search alone, s... |

1 |
ªStatistical Models of Face ImagesÐImproving Specificity,º
- Edwards, Lanitis, et al.
- 1998
(Show Context)
Citation Context ...rors. Index TermsÐAppearance models, deformable templates, model matching. 1 INTRODUCTION æ THE ªinterpretation through synthesisº approach has received considerable attention over the past few years =-=[3]-=-, [6], [11], [14]. The aim is to ªexplainº novel images by generating synthetic images that are as similar as possible, using a parameterized model of appearance. One motivation is to achieve robust s... |

1 |
ªInterpreting Face Images Using Active Appearance Models,º
- Edwards, Taylor, et al.
- 1998
(Show Context)
Citation Context ...mplexity of the algorithm is O…npixelsnmodes† at a given level. Essentially, each iteration involves sampling npixels points from the image then multiplying by a nmodes npixel matrix. In earlier work =-=[4]-=-, [1], we described a training algorithm based on regressing random displacement vectors against residual error vectors. Because of the linear assumptions being made, in the limit this would give the ... |

1 |
ªLearning Novel Views to a Single Face Image,º Proc. Second Int'l Conf. Automatic Face and Gesture Recognition,
- Vetter
- 1996
(Show Context)
Citation Context ...sÐAppearance models, deformable templates, model matching. 1 INTRODUCTION æ THE ªinterpretation through synthesisº approach has received considerable attention over the past few years [3], [6], [11], =-=[14]-=-. The aim is to ªexplainº novel images by generating synthetic images that are as similar as possible, using a parameterized model of appearance. One motivation is to achieve robust segmentation by us... |

1 |
ªActive Appearance Models,º Proc.
- Cootes, Edwards, et al.
- 1998
(Show Context)
Citation Context ...m of squares measure and essentially precomputes derivatives of the difference vector with respect to the parameters of the transformation. Hager and Belhumeur [8] describe a similar approach, but include robust kernels and models of illumination variation. Sclaroff and Isidoro [13] extend the approach to track objects which deform, modeling deformation using the low energy modes of a finite element model of the target. The approach has been used to track heads [10] using a rigid cylindrical model of the head. The Active Appearance Models described below are an extension of this approach [4], [1]. Rather than tracking a particular object, our models of appearance can match to any of a class of deformable objects (e.g., any face with any expression, rather than one persons face with a particular expression). To match rapidly, we precompute the derivatives of the residual of the match between the model and the target image and use them to compute the update steps in an iterative matching algorithm. In order to make the algorithm insensitive to changes in overall intensity, the residuals are computed in a normalized reference frame. 3 MODELING APPEARANCE Following the approach of Edwards... |

1 |
ªGeneralized Image Matching:
- Nastar, Moghaddam, et al.
- 1997
(Show Context)
Citation Context ...l to a new image. Instead, we exploit the fact that the optimization problem is similar each time so we can learn these similarities offline. This allows us to find directions of rapid convergence even though the search space has very high dimensionality. The approach has similarities with ªdifference decompositionº tracking algorithms [7], [8], [13], but rather than tracking a single deforming object we match a model which can fit a whole class of objects. 2 BACKGROUND Several authors have described methods for matching deformable models of shape and appearance to novel images. Nastar et al. [12] describe a model of shape and intensity variations using a 3D deformable model of the intensity landscape. They use a closest point surface matching algorithm to perform the fitting, which tends to be sensitive to the initialization. Lanitis et al. [11] and Edwards et al. [3] use a boundary finding algorithm (an ªActive Shape Modelº) to find the best shape, then use this to match a model of image texture. Poggio et al. [6], [9] use an optical flow algorithm to match shape and texture models iteratively, and Vetter [14] uses a general purpose optimization method to match photorealistic human f... |