## Deformotion - Deforming Motion, Shape Average and the Joint Registration and Segmentation of Images (2002)

### Cached

### Download Links

- [www.vision.cs.ucla.edu]
- [vision.ucla.edu]
- [www.vision.cs.ucla.edu]
- [www.ece.gatech.edu]
- [ftp.math.ucla.edu]
- [vision.ucla.edu]
- [www.vision.cs.ucla.edu]
- DBLP

### Other Repositories/Bibliography

Venue: | International Journal of Computer Vision |

Citations: | 108 - 15 self |

### BibTeX

@ARTICLE{Soatto02deformotion-,

author = {Stefano Soatto and Anthony J. Yezzi},

title = {Deformotion - Deforming Motion, Shape Average and the Joint Registration and Segmentation of Images},

journal = {International Journal of Computer Vision},

year = {2002},

volume = {53},

pages = {153--167}

}

### Years of Citing Articles

### OpenURL

### Abstract

What does it mean for a deforming object to be "moving" (see Fig. 1)? How can we separate the overall motion (a finite-dimensional group action) from the more general deformation (a di#eomorphism)? In this paper we propose a definition of motion for a deforming object and introduce a notion of "shape average" as the entity that separates the motion from the deformation. Our definition allows us to derive novel and e#cient algorithms to register non-equivalent shapes using region-based methods, and to simultaneously approximate and register structures in grey-scale images. We also extend the notion of shape average to that of a "moving average" in order to track moving and deforming objects through time.

### Citations

1432 |
System Identification, Theory for the User
- Ljung
- 1999
(Show Context)
Citation Context ...ssive) moving average of y. Without loss of generality one may assume that k = 1 since the difference equation above can always be reduced to first-order by augmenting the dimension of the state (see =-=[26] for details-=-). 10 (26)sγi γi gi(µ) gi(µ) µ µ γi γi Figure 3: Euclidean (top) vs. affine (bottom) registration and average. For each pair of objects γ1, γ2, the registration g1(µ), g2(µ) relative to th... |

1095 |
Active shape models – their training and application
- Cootes, Taylor, et al.
- 1994
(Show Context)
Citation Context ...ies from mathematical morphology to statistics, geology, neuroanatomy, paleontology, astronomy etc. In statistics, the study of “Shape Spaces” was championed by Kendall, Mardia and Carne among oth=-=ers [15, 20, 8, 24, 9]. Sha-=-pes are defined as the equivalence classes of N points in R M under the similarity group 2 , R MN /{SE(M) × R}. Although the framework clearly distinguishes the notion of “motion” (along the fibe... |

1090 |
Fronts propagating with curvature-dependant speed: Algorithms based on Hamilton-Jacobi formulations
- Osher, Sethian
- 1988
(Show Context)
Citation Context ...-based vs. probabilistic). For the implementation of our algorithms on deforming contours, we have chosen an implicit non-parametric representation in terms of level sets, following Osher and Sethian =-=[33]-=-, and we have implemented numerical algorithms for integrating PDEs to converge to the steady-state of an energy-based functional. However, these choices can be easily changed without altering the nat... |

1086 |
1989] Mathematical Methods of Classical Mechanics
- Arnold
(Show Context)
Citation Context ... the object at any instant of time. However, being a nonrigid object, in order to describe it at any instant of time one should really specify the trajectory of each individual particle on the sheet (=-=Arnold, 1978-=-). That is, if γ0 represents the initial collection of particles, one could provide a function f that describes how the entire set of particles evolves in time: γt = f (γ0, t). Indeed, if each particl... |

662 | Shape modeling with front propagation: a level set approach
- Malladi, Sethian, et al.
- 1995
(Show Context)
Citation Context ...s a nice taxonomy of existing work on shape and deformation and a review of the state of the art as of 1994. The variational framework has also proven very effective in the analysis of medical images =-=[28, 39, 27]-=-. Although most of the ideas are develop in a deterministic setting, many can be transposed to a probabilistic context (e.g. Zhu et al. [45]). None of these approaches, however, distinguishes a notion... |

659 | Region competition: Unifying snakes, region growing, and Bayes/MDL for multi-band image segmentation
- Zhu, Yuille
- 1996
(Show Context)
Citation Context ...en very effective in the analysis of medical images [28, 39, 27]. Although most of the ideas are develop in a deterministic setting, many can be transposed to a probabilistic context (e.g. Zhu et al. =-=[45]-=-). None of these approaches, however, distinguishes a notion of motion that is separate from the deformation; the evolution of shapes is driven by energy and regularization terms, rather than by the a... |

619 |
Active Contours
- Blake, Isard
- 1998
(Show Context)
Citation Context ...ite-dimensional group, despite infinite-dimensional deformations. Substantially different in methods, but related in the intent, is the work on stochastic filters for contour tracking and snakes (see =-=[6] and ref-=-erences therein). Our framework is designed for objects that undergo a distinct overall “global” motion while “locally” deforming. Under these assumptions, our contribution consists of a novel... |

478 |
Random Sets and Integral Geometry
- Matheron
- 1975
(Show Context)
Citation Context ...oanatomy, paleontology, astronomy etc. Some of the earlier attempts to formalize a notion of shape include D’Arcy Thompson’s treatise “Growth and Form” [38], the work of Matheron on “Stochas=-=tic Sets” [30] as -=-well as that of Thom, Giblin and others [37, 13]. In statistics, the study of “Shape Spaces” was championed by Kendall, Mardia and Carne among others [17, 24, 9, 29]. Shapes are defined as the equ... |

340 |
Axioms and fundamental equations of image processing
- Alvarez, Guichard, et al.
- 1993
(Show Context)
Citation Context ...tion, their scale-space representation at some level may be. Scale-space is a very active research area, and some of the key contributions as they relate to the material of this paper can be found in =-=[16, 36, 19, 2, 3, 1] and referen-=-ces therein. The “alignment,” or “registration,” of curves has also been used to define a notion of “shape average” by several authors (see [25] and references therein). The shape average,... |

333 | Statistical shape influence in geodesic active contours
- Leventon, Grimson, et al.
(Show Context)
Citation Context ...this paper can be found in [16, 36, 19, 2, 3, 1] and references therein. The “alignment,” or “registration,” of curves has also been used to define a notion of “shape average” by several a=-=uthors (see [25] and r-=-eferences therein). The shape average, or “prototype,” can then be used for recognition in a nearest-neighbor classification framework, or to initialize image-based segmentation by providing a “... |

301 |
The representation and matching of pictorial structures
- Fischler, Elschlager
- 1973
(Show Context)
Citation Context ...llow us to do justice to the many valuable contributions. We refer the reader to [40] for a recent survey. A common approach consists of matching collections points organized in graphs or trees (e.g. =-=[23, 12]). Belon-=-gie et al. [6] propose comparing planar contours based on their “shape context.” There, points are not bound to represent particular “landmarks” but are just a discrete representation of the c... |

225 |
General Pattern Theory
- Grenander
- 1993
(Show Context)
Citation Context ... intuition. However, Mumford has critiqued current theories of shape on the grounds that they fail to capture the essential features of perception [32]. “Deformable Templates,” pioneered by Grenan=-=der [14], do not rel-=-y on “features” or “landmarks;” rather, images are directly deformed by a (possibly infinite-dimensional) group action and compared for the best match in an “image-based” approach [44]. Th... |

212 |
Representations and knowledge in complex systems
- Grenander, Miller
- 1994
(Show Context)
Citation Context ...th that of deformation, and there is no clear distinction between the two [7]. Grenander’s work sparked a current that has been particularly successful in the analysis of medical images, for instanc=-=e [15]. We wou-=-ld like to retain some of the power and flexibility of deformable templates, but within this framework mark a clear distinction between “motion” and “deformation.” Another line of work uses va... |

197 | Geodesic active contours and level sets for the detection and tracking of moving objects
- Paragios, Deriche
(Show Context)
Citation Context ...i.e. equivalent under the group action), but they also undergo deformations. Also related to this paper is the recent work of Paragios and Deriche, where active regions are tracked as they “move.”=-= In [34]-=- the notion of motion is not made distinct from the general deformation, and therefore what is being tracked is a general (infinite-dimensional) deformation. Our aim is to define tracking as a traject... |

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

184 | Shape Manifolds, Procrustean Metrics, and Complex Projective Spaces
- Kendall
- 1984
(Show Context)
Citation Context ...the work of Matheron on “Stochastic Sets” [30] as well as that of Thom, Giblin and others [37, 13]. In statistics, the study of “Shape Spaces” was championed by Kendall, Mardia and Carne among=-= others [17, 24, 9, 29].-=- Shapes are defined as the equivalence classes of N points in R M under the similarity group, R MN /{SE(M) × R}. Shape spaces are thus organized in a fiber bundle where motion along the fibers corres... |

167 | A New Algorithm for Non-Rigid Point Matching
- Chui, Rangarajan
- 2000
(Show Context)
Citation Context ... matching is, by construction, invariant with respect to either the affine or the Euclidean group, and the resulting match is based on “features” rather than on image intensity directly, similarly=-= to [10, 11]-=-. Koenderink [22] is credited with providing some of the key ideas involved in formalizing a notion of shape that matches our intuition. However, Mumford has critiqued current theories of shape on the... |

152 | Matching shapes
- Belongie, Malik, et al.
(Show Context)
Citation Context ...any valuable contributions. We refer the reader to [32] for a recent survey. A common approach consists of matching collections of points organized in graphs or trees (e.g. [19, 11]). Belongie et al. =-=[4] propose com-=-paring planar contours based on their “shape context.” The resulting match is based on “features” rather than on image intensity directly, similar to [10]. “Deformable Templates,” pioneere... |

147 |
Structural Stability and Morphogenesis
- Thom
- 1975
(Show Context)
Citation Context ...he earlier attempts to formalize a notion of shape include D’Arcy Thompson’s treatise “Growth and Form” [38], the work of Matheron on “Stochastic Sets” [30] as well as that of Thom, Giblin=-= and others [37, 13]. In-=- statistics, the study of “Shape Spaces” was championed by Kendall, Mardia and Carne among others [17, 24, 9, 29]. Shapes are defined as the equivalence classes of N points in R M under the simila... |

123 | Computable elastic distance between shapes
- Younes
(Show Context)
Citation Context ...s and the solution of partial differential equations (PDEs) to model shape and to compute distances and similarity. In this framework, not only can the notion of alignment or distance be made precise =-=[5, 43, 31, 20, 35]-=-, but quite sophisticated theories that encompass perceptually relevant aspects, can be formalized in terms of the properties of the evolution of PDEs (e.g. [21]). The work of Kimia et al. [18] descri... |

121 |
Deformable templates for face recognition
- Yuille
- 1991
(Show Context)
Citation Context ...der [14], do not rely on “features” or “landmarks;” rather, images are directly deformed by a (possibly infinite-dimensional) group action and compared for the best match in an “image-based�=-=�� approach [44]. There, the n-=-otion of “motion” (or “alignment” or “registration”) coincides with that of deformation, and there is no clear distinction between the two [7]. Grenander’s work sparked a current that ha... |

106 |
A surface-based technique for warping threedimensional images of the brain
- Thompson, Toga
- 1996
(Show Context)
Citation Context ...s a nice taxonomy of existing work on shape and deformation and a review of the state of the art as of 1994. The variational framework has also proven very effective in the analysis of medical images =-=[28, 39, 27]-=-. Although most of the ideas are develop in a deterministic setting, many can be transposed to a probabilistic context (e.g. Zhu et al. [45]). None of these approaches, however, distinguishes a notion... |

74 | Surface learning with applications to lipreading
- Bregler, Omohundro
- 1994
(Show Context)
Citation Context ...has also proven very effective in the analysis of medical images [23, 30, 22]. Although most of these ideas are developed in a deterministic setting, many can be transposed to a probabilistic context =-=[35, 7, 31]. Sc-=-alespace is a very active research area, and some of the key contributions as they relate to the material of this paper can be found in [14, 29, 17, 1, 2] and references therein. The “alignment,” ... |

71 |
Tracking Level Sets by Level Sets: A Method for Solving Shape from Shading Problem
- Kimmel, Bruckstein
- 1995
(Show Context)
Citation Context ...s and the solution of partial differential equations (PDEs) to model shape and to compute distances and similarity. In this framework, not only can the notion of alignment or distance be made precise =-=[5, 43, 31, 20, 35]-=-, but quite sophisticated theories that encompass perceptually relevant aspects, can be formalized in terms of the properties of the evolution of PDEs (e.g. [21]). The work of Kimia et al. [18] descri... |

70 | A level set model for image classification
- Samson, Blanc-Feraud, et al.
- 2000
(Show Context)
Citation Context ...s and the solution of partial differential equations (PDEs) to model shape and to compute distances and similarity. In this framework, not only can the notion of alignment or distance be made precise =-=[5, 43, 31, 20, 35]-=-, but quite sophisticated theories that encompass perceptually relevant aspects, can be formalized in terms of the properties of the evolution of PDEs (e.g. [21]). The work of Kimia et al. [18] descri... |

69 | Stereoscopic segmentation
- Yezzi, Soatto
- 2001
(Show Context)
Citation Context ...art of the inference problem. Errors in the pre-processing stage can never be compensated, and the “registration” cannot be adapted as the 3soriginal contours evolve towards their average. Similar=-=ly, [41]-=- performs the joint segmentation of a number of images by assuming that their registration (stereo calibration) is given. We wish to extend these approaches to situations where the calibration/registr... |

66 | shocks, and deformations I: The components of two-dimensional shape and the reaction-diffusion space,” Intern
- Kimia, Tannenbaum, et al.
- 1995
(Show Context)
Citation Context ...31, 20, 35], but quite sophisticated theories that encompass perceptually relevant aspects, can be formalized in terms of the properties of the evolution of PDEs (e.g. [21]). The work of Kimia et al. =-=[18] describ-=-es a scale-space that corresponds to various stages of evolution of a diffusing PDE, and a “reacting” PDE that splits “salient parts” of planar contours by generating singularities. [18] also ... |

63 |
Mathematical theories of shape: Do they model perception
- Mumford
- 1991
(Show Context)
Citation Context ...ved in formalizing a notion of shape that matches our intuition. However, Mumford has critiqued current theories of shape on the grounds that they fail to capture the essential features of perception =-=[32]. “Deforma-=-ble Templates,” pioneered by Grenander [14], do not rely on “features” or “landmarks;” rather, images are directly deformed by a (possibly infinite-dimensional) group action and compared for... |

61 |
State of the art in shape matching
- Veltkamp, C, et al.
(Show Context)
Citation Context ...rature exists for the problem of “matching” or “aligning” objects based on their images, and space limitations do not allow us to do justice to the many valuable contributions. We refer the re=-=ader to [40] f-=-or a recent survey. A common approach consists of matching collections points organized in graphs or trees (e.g. [23, 12]). Belongie et al. [6] propose comparing planar contours based on their “shap... |

54 | Scale-space properties of the multiscale morphological dilation-erosion
- Jackway, Deriche
- 1996
(Show Context)
Citation Context ...tion, their scale-space representation at some level may be. Scale-space is a very active research area, and some of the key contributions as they relate to the material of this paper can be found in =-=[16, 36, 19, 2, 3, 1] and referen-=-ces therein. The “alignment,” or “registration,” of curves has also been used to define a notion of “shape average” by several authors (see [25] and references therein). The shape average,... |

48 |
The Riemannian structure of euclidean shape spaces: a novel environment for statistics
- Kendall
- 1993
(Show Context)
Citation Context ...the work of Matheron on “Stochastic Sets” [30] as well as that of Thom, Giblin and others [37, 13]. In statistics, the study of “Shape Spaces” was championed by Kendall, Mardia and Carne among=-= others [17, 24, 9, 29].-=- Shapes are defined as the equivalence classes of N points in R M under the similarity group, R MN /{SE(M) × R}. Shape spaces are thus organized in a fiber bundle where motion along the fibers corres... |

39 | Intrinsic scale space for images on surfaces: The geodesic curvature flow
- Kimmel
- 1997
(Show Context)
Citation Context ...tion, their scale-space representation at some level may be. Scale-space is a very active research area, and some of the key contributions as they relate to the material of this paper can be found in =-=[16, 36, 19, 2, 3, 1] and referen-=-ces therein. The “alignment,” or “registration,” of curves has also been used to define a notion of “shape average” by several authors (see [25] and references therein). The shape average,... |

38 |
On Growth And
- Thompson
- 1942
(Show Context)
Citation Context ...ematical morphology to statistics, geology, neuroanatomy, paleontology, astronomy etc. Some of the earlier attempts to formalize a notion of shape include D’Arcy Thompson’s treatise “Growth and =-=Form” [38], the wo-=-rk of Matheron on “Stochastic Sets” [30] as well as that of Thom, Giblin and others [37, 13]. In statistics, the study of “Shape Spaces” was championed by Kendall, Mardia and Carne among other... |

37 | A scale-space approach to nonlocal optical flow calculations
- Alvarez, Weickert, et al.
- 1999
(Show Context)
Citation Context ...any can be transposed to a probabilistic context [35, 7, 31]. Scalespace is a very active research area, and some of the key contributions as they relate to the material of this paper can be found in =-=[14, 29, 17, 1, 2] and referen-=-ces therein. The “alignment,” or “registration,” of curves has also been used to define a notion of “shape average” by several authors (see [21] and references therein). The shape average,... |

36 | Graphs, surfaces and homology - Giblin - 1981 |

30 |
A geometric approach to segmentation and analysis of 3-D medical images
- Malladi, Kimmel, et al.
- 1996
(Show Context)
Citation Context ...s a nice taxonomy of existing work on shape and deformation and a review of the state of the art as of 1994. The variational framework has also proven very effective in the analysis of medical images =-=[28, 39, 27]-=-. Although most of the ideas are develop in a deterministic setting, many can be transposed to a probabilistic context (e.g. Zhu et al. [45]). None of these approaches, however, distinguishes a notion... |

29 |
Shape Distributions for Landmark Data
- Mardia, Dryden
- 1989
(Show Context)
Citation Context ...the work of Matheron on “Stochastic Sets” [30] as well as that of Thom, Giblin and others [37, 13]. In statistics, the study of “Shape Spaces” was championed by Kendall, Mardia and Carne among=-= others [17, 24, 9, 29].-=- Shapes are defined as the equivalence classes of N points in R M under the similarity group, R MN /{SE(M) × R}. Shape spaces are thus organized in a fiber bundle where motion along the fibers corres... |

27 |
Group action, diffeomorphism and matching: A general framework
- Miller, Younes
- 1999
(Show Context)
Citation Context |

24 | Constructing 2d curve atlases - Sebastian, Crisco, et al. - 2000 |

21 |
The geometry of shape spaces
- Carne
- 1990
(Show Context)
Citation Context ...ies from mathematical morphology to statistics, geology, neuroanatomy, paleontology, astronomy etc. In statistics, the study of “Shape Spaces” was championed by Kendall, Mardia and Carne among oth=-=ers [15, 20, 8, 24, 9]. Sha-=-pes are defined as the equivalence classes of N points in R M under the similarity group 2 , R MN /{SE(M) × R}. Although the framework clearly distinguishes the notion of “motion” (along the fibe... |

21 | Multivalued distance maps for motion planning on surfaces with moving obstacles
- Kimmel, Kiryati, et al.
- 1998
(Show Context)
Citation Context ...istance be made precise [5, 43, 31, 20, 35], but quite sophisticated theories that encompass perceptually relevant aspects, can be formalized in terms of the properties of the evolution of PDEs (e.g. =-=[21]). The w-=-ork of Kimia et al. [18] describes a scale-space that corresponds to various stages of evolution of a diffusing PDE, and a “reacting” PDE that splits “salient parts” of planar contours by gene... |

18 |
Morphological approach to multiscale analysis: From principles to equations
- Alvarez, Morel
- 1994
(Show Context)
Citation Context ...any can be transposed to a probabilistic context [35, 7, 31]. Scalespace is a very active research area, and some of the key contributions as they relate to the material of this paper can be found in =-=[14, 29, 17, 1, 2] and referen-=-ces therein. The “alignment,” or “registration,” of curves has also been used to define a notion of “shape average” by several authors (see [21] and references therein). The shape average,... |

12 | A distance for elastic matching in object recognition
- Azencott, Coldefy, et al.
- 1996
(Show Context)
Citation Context ...s and the solution of partial differential equations (PDEs) to model shape and to compute distances and similarity. In this framework, not only can the notion of alignment or distance be made precise =-=[3, 33, 25, 28]-=-, but quite sophisticated theories that encompass perceptually relevant aspects can be formalized in terms of the properties of the evolution of PDEs (e.g. [18, 16]). Kimia et al. [16] describes a sca... |

7 |
Scale-space Theory in Computer Vision
- Romeny, Florack, et al.
- 1997
(Show Context)
Citation Context |

5 |
Motion detection in meteorological images sequences: Two methods and their comparison
- Bereziat, Herlin, et al.
- 1997
(Show Context)
Citation Context ...the best match in an “imagebased” approach [34]. There, the notion of “motion” (or “alignment” or “registration”) coincides with that of deformation, and there is no clear distinction =-=between the two [5]-=-. Another line of work uses variational methods and the solution of partial differential equations (PDEs) to model shape and to compute distances and similarity. In this framework, not only can the no... |

5 |
A variational approach to joint segmentation and registration
- Yezzi, Zollei, et al.
- 2001
(Show Context)
Citation Context ...ming that their registration (stereo calibration) is given. We wish to extend these approaches to situations where the calibration/registration is not known a-priori. A somewhat complementary work is =-=[42]-=-, where objects, assumed to be identical except for a group action, are registered by minimizing a regionbased cost functional. We wish to extend that approach to situations where the objects are not ... |

3 |
Distortion invariant object rcognition in the dynamic link architecture
- Lades, Borbruggen, et al.
- 1993
(Show Context)
Citation Context ...llow us to do justice to the many valuable contributions. We refer the reader to [40] for a recent survey. A common approach consists of matching collections points organized in graphs or trees (e.g. =-=[23, 12]). Belon-=-gie et al. [6] propose comparing planar contours based on their “shape context.” There, points are not bound to represent particular “landmarks” but are just a discrete representation of the c... |

2 |
Corpus callosum shape analysis: A comparative study of group differences associated with dyslexia, Au: Pls. update. gender and handedness
- Dutta, Jain
- 2001
(Show Context)
Citation Context ... matching is, by construction, invariant with respect to either the affine or the Euclidean group, and the resulting match is based on “features” rather than on image intensity directly, similarly=-= to [10, 11]-=-. Koenderink [22] is credited with providing some of the key ideas involved in formalizing a notion of shape that matches our intuition. However, Mumford has critiqued current theories of shape on the... |