C. Nastar and N. Ayache. Fast segmentation, track- ing, and analysis of deformable objects. In IEEE Proceedings of the Third International Conference on Computer Vision (ICCV'93), Berlin, May 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....period of the system [11] given by: T 0 r K (3) K c m T Therefore, if we want to increase stiffness, we have to decrease Deltat below the new decreased value of T 0 . Our model can be reduced to this case provided the natural lengths of the springs are supposed equal to zero [12]. For a same animation time, the number of iterations needed will then be greater, and the algorithm will be more costly In our model, all springs have a natural length non equal to zero, but they remain nevertheless intrinsically linear springs. However, these springs once coupled lead our ....

Nastar Chahab, Ayache Nicholas. Fast Segmentation, Tracking, and Analysis of Deformable Objects. In Proceedings of the Fourth International Conference on Computer Vision (ICCV '93), Berlin, 1993.

....improving the global control of FFD have been devised. A rst approach consists of decomposing a deformation eld into a set of hierarchical deformation modes. For instance, in [11, 19] Metaxas and Terzopoulos superimpose a deformable superquadrics over a surface spline function. Modal analysis [13, 17], Fourier domain analysis [16] or wavelet basis [21] provide a set of deformation modes with a decreasing scale of extent. Principal component analysis [3] introduces meaningful deformation modes from statistical study of shape variation. A second approach consists of including global parameters ....

C. Nastar and N. Ayache. Fast segmentation, tracking, and analysis of deformable objects. In Proceedings of the Fourth International Conference on Computer Vision (ICCV 93), Berlin, May 1993. also in SPIE, Geometric Methods in Computer Vision, San-Diego, 1993.

....i.e. as the surface (x, y, I(x, y) as shown, for example, in Figure 1 and developed an efficient method to warp one image onto another using a physically based deformation model. In this section we briefly review the mathematics of this approach (for further details the reader is referred to [11, 13, 12]) The intensity surface is modeled as a deformable mesh and is governed by Lagrangian dynamics [1] MI CO KU F(t) 1) where U [ Axi, Ayi, Azi, T is a vector storing nodal displacements, M, C and K are respectively the mass, damping and stiffness matrices of the system, and F ....

....to a specific case of our formulation where the closest point Pi has to have the same intensity as Mi i.e. MiPi is parlcl to the XY plane. Wc do not make that assumption here. Solutions of the governing equation arc typically obtMncd using an cigcnvcctor bascd modal decomposi tion [14, 11, 10]. In particular, the vibration modes 0(i) of the previous deformable surface are the vector solutions of the eigenproblem: where (i) is the i th eigenfrequency of the system. Solving the governing equations in the modM basis leads to scMar equations where the unknown i(i) is the amplitude of mode ....

C. Nastar and N. Ayache. Fast segmentation, track- ing, and analysis of deformable objects. In IEEE Proceedings of the Third International Conference on Computer Vision (ICCV'93), Berlin, May 1993.

....we introduce deformable intensity surfaces for object matching and recognition. Our method allows to deal effectively with object appearance variations. The basic idea is to warp the intensity surface of the test image into the intensity surface of the probe image by using a physics based model [14, 13], and then measure the low order strain energy of the deformable surface for recognition. We further show how to tailor the space of allowable XYI deformations to fit the actual variation found in individual target classes. This is accomplished by a statistical analysis of observed image to image ....

....1: An image and its XYI representation 3 Intensity surface matching Many surface matching techniques are available in the computer vision literature. For instance, we can point out geometric surface matching based on differential features [6] and the dual approach of physicallybased matching [14]. As we shall see, physically based matching has a nice compression ability. 3.1 Surface modeling In this section, we model the intensity surface S in order to deform it to the intensity surface S . The proposed approach is described in [14] Intensity surface S is modeled by a deformable mesh ....

[Article contains additional citation context not shown here]

C. Nastar and N. Ayache. Fast segmentation, tracking, and analysis of deformable objects. In Proceedings of the Fourth International Conference on Computer Vision (ICCV '93), Berlin, May 1993.

.... physicallybased deformable models [11, 5, 4] Among these works, modal analysis, first introduced in computer vision by Pentland et al. 9] has the advantage of being a frequency based technique in which nonrigid motion is expressed in the free vibrations basis (modes) of the deformable object [7, 6, 8]. More recently, many researchers study the temporal evolution of deformable models in order to analyze time sequences of images [10, 3, 1] In this paper, we propose a unified approach for nonrigid motion estimation from time sequences of multidimensional images by taking into account both the ....

....motion. Our method has important implications in automatic diagnosis of heart diseases and in 4D data compression. 2 An elastically deformable model A discrete mass spring mesh of N nodes can be elastically deformed in images to track the contour of the deforming object (e.g. the left ventricle) [7]. More precisely, in 3D, the deformation of the system is governed by the 3N dimensional differential matrix equation [2] M U C U KU = F(t) 1) where U is a vector storing nodal displacements M, C and K are respectively the mass, damping and stiffness matrices of the system, and F is the ....

Chahab Nastar and Nicholas Ayache. Fast segmentation, tracking, and analysis of deformable objects. In Proceedings of the Fourth International Conference on Computer Vision (ICCV '93), Berlin, May 1993.

....we introduce deformable intensity surfaces for object matching and recognition. Our method allows to deal effectively with object appearance variations. The basic idea is to warp the intensity surface of the test image into the intensity surface of the probe image by using a physics based model [14, 13], and then measure the low order strain energy of the deformable surface for recognition. We further show how to tailor the space of allowable XY I deformations to fit the actual variation found in individual target classes. This is accomplished by a statistical analysis of observed ....

....1: An image and its XY I representation 3 Intensity surface matching Many surface matching techniques are available in the computer vision literature. For instance, we can point out geometric surface matching based on differential features [6] and the dual approach of physicallybased matching [14]. As we shall see, physically based matching has a nice compression ability. 3.1 Surface modeling In this section, we model the intensity surface S in order to deform it to the intensity surface S 0 . The proposed approach is described in [14] Intensity surface S is modeled by a deformable ....

[Article contains additional citation context not shown here]

C. Nastar and N. Ayache. Fast segmentation, tracking, and analysis of deformable objects. In Proceedings of the Fourth International Conference on Computer Vision (ICCV '93), Berlin, May 1993.

....comparison and the static method of shape comparison using Fourier shape descriptors (see [18] for a review) However, the 3D case developed by our model cannot be easily described by those considerations. 2 An elastically deformable model Let us consider a discrete mass spring mesh of N nodes [25, 9, 15] . Using the equations of dynamics, such a structure can be elastically deformed in 2D or 3D images to match the contour of an object of interest. Now if we take a sequence of images showing the nonrigidly moving object, the model can achieve both segmentation and tracking of the object surface ....

Chahab Nastar and Nicholas Ayache. Fast segmentation, tracking, and analysis of deformable objects. In Proceedings of the Fourth International Conference on Computer Vision (ICCV '93), Berlin, May 1993.

....the approach is demonstrated by a set of experimental results on 3D medical data. 1 Background Following the theory of deformable models, physically based modelling for nonrigid motion analysis has become extremely popular [5] For purposes of deformation analysis, we make use of modal analysis [4, 2], a well known mechanical engineering technique which consists in decomposing and approximating the motion in the free vibrations basis (modes) of the model. 2 Modal analysis for deformable models Consider a discrete mass spring mesh of N nodes. Using the equations of dynamics, such a structure ....

.... matrix equation above has the nice property of decoupling into three N order matrix equations in each space direction, as it clearly appears in equation (5) From now on, we will keep this assumption which reduces computational cost as nodal vectors and matrices are of order N instead of 3N [2]. Free vibrations of a one dimensional lattice : Consider a set of ions distributed along a chain at points separated by a distance a, so that the lattice vectors are R = na for n 2 f1; Ng. If only neighboring ions interact, we may take the harmonic potential energy to have the form : V ....

Chahab Nastar and Nicholas Ayache. Fast segmentation, tracking, and analysis of deformable objects. In Proceedings of the Fourth International Conference on Computer Vision (ICCV '93), Berlin, May 1993.

....is quite related to the surface deformation methods because we deform the surfaces in order to bring the corresponding points nearer. For example, the deformation techniques presented in [BK89] Inria Rigid, Affine and Locally Affine Registration of Free Form Surfaces 5 [MT93] MT91] PS91] or [NA93] are very interesting. These authors deform the surfaces using a physical model involving internal and external forces. Our deformations are quite different: they are the result of a geometric transformation, and the constraints used are based on geometric differential informations. Hence, unlike ....

Chahab Nastar and Nicholas Ayache. Fast segmentation, tracking, and analysis of deformable objects. In Proceedings of the Fourth International Conference on Computer Vision (ICCV '93), Berlin, May 1993.

....formulations in object modeling and recognition which employ the principal modes or characteristic degrees of freedom for description. The identification and parametric representation of a system in terms of these principal modes is at the core of recent advances in physically based modeling [20, 17] and parametric descriptions of shape [7, 2, 11] On the other hand, view based eigentechniques have recently provided some of the best results in object recognition [21, 19] In this paper, we propose a new method which combines both the physically based modes of vibration and the ....

....distance maps created at step 2. Note that steps 1 to 4 are pre processing steps. Steps 1 and 2 provide respectively intensity and spatial smoothing x I(x) S S Figure 1: Intensity surface S being pulled towards S 0 by image forces of the image. The dynamic process of step 5 is described in [17] ; to sum up, the intensity surface S is modeled as a deformable mesh of size N = n Theta n 0 nodes, ruled by Lagrangian dynamics : M U C U KU = F(t) 1) where U = Deltax i ; Deltay i ; Deltaz i ; T is a vector storing nodal displacements, M, C and K are respectively the ....

[Article contains additional citation context not shown here]

C. Nastar and N. Ayache. Fast segmentation, tracking, and analysis of deformable objects. In Proceedings of the Fourth International Conference on Computer Vision (ICCV '93), Berlin, May 1993.

....surface of the test image into the intensity surface of the probe image by using a physics based model, and then measure the low order strain energy of the deformable surface for recognition. This recognition metric has to do with the eigenmodes of the intensity surface modeled as a thinplate [18, 17]. We choose face recognition as an application for our method. For over 20 years, automatic face recognition has been the center of attention of computer vision researchers. Face recognition is indeed a difficult object recognition problem, because of the wide variation in the appearance of a ....

....3. 1 Theoretical background Following the theory of active contour models [11, 26] several models have been developed that deal explicitly with deformable surfaces, among them : deformable superquadrics [19, 25] surface snakes [6, 13] particle systems [24] splines [3] and elastic thin plates [21, 18]. The above models usually evolve in Euclidean 3D space, however, deformable templates which evolve in XY I space with application to feature extraction have been investigated by Yuille et al. [29] Hence, deformable intensity surfaces is a new approach to matching and recognition. The ....

[Article contains additional citation context not shown here]

C. Nastar and N. Ayache. Fast segmentation, tracking, and analysis of deformable objects. In Proceedings of the Fourth International Conference on Computer Vision (ICCV '93), Berlin, May 1993.

....details the reader 1 In fact, by simply disabling the I component of our deformations we can obtain a standard 2D deformable mesh which yields correspondences similar to an optical flow technique with thin plate regularizers. Figure 1: An image and its XYI surface representation is referred to [11, 12]) The intensity surface is modeled as a deformable mesh and is governed by Lagrangian dynamics [1] M U C U KU = F(t) 2) where U = Deltax i ; Deltay i ; Deltaz i ; T is a vector storing nodal displacements, M, C and K are respectively the mass, damping and stiffness ....

....of our formulation where the closest point P i has to have the same intensity as M i i.e. Gamma Gamma Gamma M i P i is parallel to the XY plane. We do not make that assumption here. Solutions of the governing equation are typically obtained using an eigenvector based modal decomposition [13, 11, 10]. In particular, the vibration modes OE(i) of the previous deformable surface are the vector solutions of the eigenproblem : KOE = 2 MOE (4) where (i) is the i th eigenfrequency of the system. Solving the governing equations in the modal basis leads to scalar equations where the unknown ....

C. Nastar and N. Ayache. Fast segmentation, tracking, and analysis of deformable objects. In IEEE Proceedings of the Third International Conference on Computer Vision (ICCV'93), Berlin, May 1993.

....i.e. as the surface (x; y; I(x;y) as shown, for example, in Figure 1 and developed an efficient method to warp one image onto another using a physically based deformation model. In this section we briefly review the mathematics of this approach (for further details the reader is referred to [11, 13, 12]) The intensity surface is modeled as a deformable mesh and is governed by Lagrangian dynamics [1] M U C U KU = F(t) 1) where U = Deltax i ; Deltay i ; Deltaz i ; T is a vector storing nodal displacements, M, C and K are respectively the mass, damping and stiffness ....

....of our formulation where the closest point P i has to have the same intensity as M i i.e. Gamma Gamma Gamma M i P i is parallel to the XY plane. We do not make that assumption here. Solutions of the governing equation are typically obtained using an eigenvector based modal decomposition [14, 11, 10]. In particular, the vibration modes OE(i) of the previous deformable surface are the vector solutions of the eigenproblem : KOE = 2 MOE (3) where (i) is the i th eigenfrequency of the system. Solving the governing equations in the modal basis leads to scalar equations where the unknown ....

C. Nastar and N. Ayache. Fast segmentation, tracking, and analysis of deformable objects. In IEEE Proceedings of the Third International Conference on Computer Vision (ICCV'93), Berlin, May 1993.

No context found.

C. Nastar and N. Ayache, Fast Segmentation, Tracking, and Analysis of Deformable Objects, Proc. ICCV '93, 275-279, (1993).

No context found.

C. Nastar and N. Ayache. Fast segmentation, tracking, and analysis of deformable objects. In Proc. ICCV, pages 275--279, 1993.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC