## Expression-invariant 3D face recognition (2003)

### Cached

### Download Links

- [www.cs.technion.ac.il]
- [www.cs.technion.ac.il]
- [www.face-rec.org]
- [visl.technion.ac.il]
- [vista.eng.tau.ac.il]
- DBLP

### Other Repositories/Bibliography

Citations: | 80 - 17 self |

### BibTeX

@MISC{Bronstein03expression-invariant3d,

author = {Alexander M. Bronstein and Michael M. Bronstein and Ron Kimmel},

title = {Expression-invariant 3D face recognition},

year = {2003}

}

### Years of Citing Articles

### OpenURL

### Abstract

We present a novel 3D face recognition approach based on geometric invariants introduced by Elad and Kimmel. The key idea of the proposed algorithm is a representation of the facial surface, invariant to isometric deformations, such as those resulting from different expressions and postures of the face. The obtained geometric invariants allow mapping 2D facial texture images into special images that incorporate the 3D geometry of the face. These signature images are then decomposed into their principal components. The result is an efficient and accurate face recognition algorithm that is robust to facial expressions. We demonstrate the results of our method and compare it to existing 2D and 3D face recognition algorithms.

### Citations

1503 | Eigenfaces vs. Fisherfaces: Recognition using class specific linear projection
- Belhumeur, Hespanha, et al.
- 1997
(Show Context)
Citation Context ...face imaging [2]. This has become known as the eigenface algorithm and is now a golden standard in face recognition. Later, algorithms inspired by eigenfaces that use similar ideas were proposed (see =-=[3]-=-, [4], [5]). However, all the 2D (image-based) face recognition methods appear to be sensitive to illuminations conditions, head orientations, facial expressions and makeup. These limitations of 2D me... |

941 |
Face recognition using eigenface
- Turk, Petland
- 1991
(Show Context)
Citation Context ...nd mouth) and their geometric relationships (angles, length, ratios, etc.). In 1991, Turk and Pentland introduced the revolutionary idea of applying principal component analysis (PCA) to face imaging =-=[2]-=-. This has become known as the eigenface algorithm and is now a golden standard in face recognition. Later, algorithms inspired by eigenfaces that use similar ideas were proposed (see [3], [4], [5]). ... |

762 | A Morphable Model for the Synthesis of 3D Faces
- Blanz, Vetter
- 1999
(Show Context)
Citation Context ...ed light range camera [7]. This approach demonstrated better robustness to head orientations. Another attempt to cope with the problem of head pose using 3D morphable head models is presented in [8], =-=[9]-=-. Mavridis et al. incorporated a range map of the face into the classical face recognition algorithms based on PCA and hidden Markov models [10]. Particularly, this approach showed robustness to large... |

380 |
Modern Multidimensional Scaling - Theory and Applications
- Borg, Groenen
- 1997
(Show Context)
Citation Context ...he obtained m-dimensional representation is a set of points xi ∈ R m (i = 1,…,n), corresponding to the surface points pi. Different MDS methods can be derived using different embedding error criteria =-=[19]-=-. A particular case is the classical scaling, introduced by Young and Householder [20]. The embedding in R m is performed by double-centering the matrix ∆ B=− J∆ J. (5) 1 2 (here J = I - U; I is a n×... |

114 |
Face Recognition
- Shakhnarovich, Moghaddam
- 2004
(Show Context)
Citation Context ...ng [2]. This has become known as the eigenface algorithm and is now a golden standard in face recognition. Later, algorithms inspired by eigenfaces that use similar ideas were proposed (see [3], [4], =-=[5]-=-). However, all the 2D (image-based) face recognition methods appear to be sensitive to illuminations conditions, head orientations, facial expressions and makeup. These limitations of 2D methods stem... |

103 | Three-dimensional face recognition
- Bronstein, Bronstein, et al.
- 2004
(Show Context)
Citation Context ...ne of the primary limiting factors of other 3D face recognition algorithms as well. In this work, we present a geometric framework for efficient and accurate face recognition using 3D data (patented, =-=[11]-=-). Our method is based on geometric invariants of the human face and performs a non-rigid surface comparison, allowing deformations, typical to the human face due to facial expressions. 2 Non-rigid Su... |

99 |
Discussion of a set of points in terms of their mutual distances
- Young, Householder
- 1938
(Show Context)
Citation Context ...esponding to the surface points pi. Different MDS methods can be derived using different embedding error criteria [19]. A particular case is the classical scaling, introduced by Young and Householder =-=[20]-=-. The embedding in R m is performed by double-centering the matrix ∆ B=− J∆ J. (5) 1 2 (here J = I - U; I is a n×n identity matrix, and U is a matrix consisting entirely of ones). The first m eigenve... |

93 | Content based retrieval of vrml objects: an iterative and interactive approach
- Elad, Tal, et al.
- 2001
(Show Context)
Citation Context ..., to allow comparison between canonical forms, they must be aligned. This is possible, for example, by setting the first-order moments (center of mass) and the mixed second-order moments to zero (see =-=[21]-=-). 2.2 Measuring Geodesic Distances on Triangulated Manifolds One of the crucial steps in the construction of the canonical form of a given surface, is an efficient algorithm for the computation of th... |

89 | Texture mapping using surface flattening via multidimensional scaling
- Zigelman, Kimmel, et al.
(Show Context)
Citation Context ...dels. In their pioneering work, they applied an MDS technique to flatten convoluted cortical surfaces of the brain, onto a plane, in order to study their functional architecture [15]. Zigelman et al. =-=[16]-=- and Grossman et al. [17] extended some of these ideas to the problem of texture mapping and voxelbased cortex flattening. A generalization of this approach was introduced in the recent work of Elad a... |

67 |
A numerical solution to the generalized mapmaker’s problem: flattening nonconvex polyhedral surfaces
- Schwartz, Shaw, et al.
- 1989
(Show Context)
Citation Context ... surfaces by planar models. In their pioneering work, they applied an MDS technique to flatten convoluted cortical surfaces of the brain, onto a plane, in order to study their functional architecture =-=[15]-=-. Zigelman et al. [16] and Grossman et al. [17] extended some of these ideas to the problem of texture mapping and voxelbased cortex flattening. A generalization of this approach was introduced in the... |

63 |
M.: A 3-D Recognition and Positioning Algorithm Using Geometrical Matching Between Primitive Surfaces
- Faugeras, Hebert
(Show Context)
Citation Context ...ssions. 2 Non-rigid Surface Matching Classical surface matching methods, based on finding a Euclidean transformation of two surfaces which maximizes some shape similarity criterion (see, for example, =-=[12]-=-, [13], [14]), are suitable mainly for rigid objects. Human face can not be considered a rigid object since it undergoes deformations resulting from facial expressions. On the other hand, the class of... |

42 | Toward optimal structured light patterns
- Horn, Kiryati
- 1999
(Show Context)
Citation Context ...struction. One of the most popular approaches known as structured light, is based on projecting a pattern on the object surface and extracting the object geometry from the deformations of the pattern =-=[24]-=-. A more robust and accurate version of this approach uses a series of black and white stripes projected sequentially and is known as coded light. The patterns form a binary code, that allows the reco... |

38 |
The model method in facial recognition
- Bledsoe
- 1964
(Show Context)
Citation Context ...rics, is non-intrusive and can be used even without the subject’s knowledge. State-of-the-art face recognition systems are based on a 40-year heritage of 2D algorithms, dating back to the early 1960s =-=[1]-=-. The first face recognition methods used the geometry of key points (like the eyes, nose and mouth) and their geometric relationships (angles, length, ratios, etc.). In 1991, Turk and Pentland introd... |

37 |
Bending invariant representations for surfaces
- Elad, Kimmel
- 2001
(Show Context)
Citation Context ...man et al. [17] extended some of these ideas to the problem of texture mapping and voxelbased cortex flattening. A generalization of this approach was introduced in the recent work of Elad and Kimmel =-=[18]-=-, as a framework for object recognition. They introduced an efficient algorithm to construct a signature for isometric surfaces. This method, referred to as bending-invariant canonical forms, is the c... |

29 | Mixtures of local linear subspaces for face recognition
- Frey, Colmenarez, et al.
- 1998
(Show Context)
Citation Context ...imaging [2]. This has become known as the eigenface algorithm and is now a golden standard in face recognition. Later, algorithms inspired by eigenfaces that use similar ideas were proposed (see [3], =-=[4]-=-, [5]). However, all the 2D (image-based) face recognition methods appear to be sensitive to illuminations conditions, head orientations, facial expressions and makeup. These limitations of 2D methods... |

27 |
A review of the theory, algorithms, and applications of level set method for propagating surfaces. Acta numerica
- Sethian
- 1996
(Show Context)
Citation Context ... henceforth referred to as fast marching on triangulated domains (FMTD), was used by Elad and Kimmel [18]. FMTD was proposed by Kimmel and Sethian [22] as a generalization of the fast marching method =-=[23]-=-. Using FMTD, the geodesic distances between a surface vertex and the rest of the n surface vertices can be computed in O(n) operations. We use this method for the bending invariant canonical form com... |

25 |
Computing geodesic on manifolds
- Kimmel, Sethian
- 1998
(Show Context)
Citation Context ...thm for distance computation on triangulated domains, henceforth referred to as fast marching on triangulated domains (FMTD), was used by Elad and Kimmel [18]. FMTD was proposed by Kimmel and Sethian =-=[22]-=- as a generalization of the fast marching method [23]. Using FMTD, the geodesic distances between a surface vertex and the rest of the n surface vertices can be computed in O(n) operations. We use thi... |

17 | Automatic face authentication from 3D surface
- Beumier, Acheroy
- 1998
(Show Context)
Citation Context ...pp. 62-70, 2003. © Springer-Verlag Berlin Heidelberg 2003sExpression-Invariant 3D Face Recognition 63 tral and lateral profiles from the 3D facial surface, acquired by a structured light range camera =-=[7]-=-. This approach demonstrated better robustness to head orientations. Another attempt to cope with the problem of head pose using 3D morphable head models is presented in [8], [9]. Mavridis et al. inco... |

17 | Face recognition using component-based SVM. Classification and morphable models
- Huang, Blanz, et al.
(Show Context)
Citation Context ...uctured light range camera [7]. This approach demonstrated better robustness to head orientations. Another attempt to cope with the problem of head pose using 3D morphable head models is presented in =-=[8]-=-, [9]. Mavridis et al. incorporated a range map of the face into the classical face recognition algorithms based on PCA and hidden Markov models [10]. Particularly, this approach showed robustness to ... |

17 | Computational surface flattening: a voxel-based approach
- Grossman, Kiryati, et al.
(Show Context)
Citation Context ... work, they applied an MDS technique to flatten convoluted cortical surfaces of the brain, onto a plane, in order to study their functional architecture [15]. Zigelman et al. [16] and Grossman et al. =-=[17]-=- extended some of these ideas to the problem of texture mapping and voxelbased cortex flattening. A generalization of this approach was introduced in the recent work of Elad and Kimmel [18], as a fram... |

15 | Face recognition from frontal and profile views
- Gordon
- 1997
(Show Context)
Citation Context ...ation is viewpoint- and lighting-condition independent, i.e. lacks the “intrinsic” weaknesses of 2D approaches. Gordon showed that combining frontal and profile views can improve recognition accuracy =-=[6]-=-. This idea was extended by Beumier and Acheroy, who compared cenJ. Kittler and M.S. Nixon (Eds.): AVBPA 2003, LNCS 2688, pp. 62-70, 2003. © Springer-Verlag Berlin Heidelberg 2003sExpression-Invariant... |

10 |
The HISCORE face recognition application: Affordable desktop face recognition based on a novel 3D camera
- Mavridis, Tsalakanidou, et al.
- 2001
(Show Context)
Citation Context ...pose using 3D morphable head models is presented in [8], [9]. Mavridis et al. incorporated a range map of the face into the classical face recognition algorithms based on PCA and hidden Markov models =-=[10]-=-. Particularly, this approach showed robustness to large variations in color and illumination and use of cosmetics, and also allowed separating the face from cluttered background. However, none of the... |

5 |
The free form matching problem. Machine vision for threedimensional scene
- Besl
- 1990
(Show Context)
Citation Context .... 2 Non-rigid Surface Matching Classical surface matching methods, based on finding a Euclidean transformation of two surfaces which maximizes some shape similarity criterion (see, for example, [12], =-=[13]-=-, [14]), are suitable mainly for rigid objects. Human face can not be considered a rigid object since it undergoes deformations resulting from facial expressions. On the other hand, the class of trans... |

5 |
Recovering the position and orientation of free-form objects from image contours using 3D distance map
- Barequet, Sharir
- 1997
(Show Context)
Citation Context ...n-rigid Surface Matching Classical surface matching methods, based on finding a Euclidean transformation of two surfaces which maximizes some shape similarity criterion (see, for example, [12], [13], =-=[14]-=-), are suitable mainly for rigid objects. Human face can not be considered a rigid object since it undergoes deformations resulting from facial expressions. On the other hand, the class of transformat... |