Sclaroff, S., and Pentland, A., A modal framework for correspondence and description. In Proc. of IEEE, pp. 308-313, 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....in the presence of noise and occlusion. It establishes an explicit correspondence between the template and image point sets, and the strength of this relationship can be binary or soft in the sense of representing a confidence in the given match. Among the previous work done on related problems: [Sclaroff and Pentland, 1993] describe a matching technique for correspondence and recognition in which the object is described in terms of its eigenmodes, and similarities between objects are measured in terms of modal deformation energies. A limitation of this technique is in its ability to handle partial matches or to ....

Sclaroff, S. and Pentland, A. (1993). A modal framework for correspondence and description. In IEEE Conference on Computer Vision and Pattern Recognition, pages 308--313. IEEE Press.

....signi#cant modes. 2. 3 The Finite Element Method The Finite Element Method (FEM) is an engineering technique for ef#cient computational simulation of physical systems (see, for example, Bathe [25] Pentland and Sclaroff describe the application of these techniques to problems in computer vision [26, 27, 28, 29]. The approach taken is to build an elastic physical model of a deformable object and use #nite element analysis 15 to produce a compact, orthogonal set of shape parameters suitable for tracking and recognition tasks. Nastar and Ayache have successfully applied these techniques in the analysis ....

S Sclaroff and A Pentland. A modal framework for correspondence and description. In Proc 4th International Conference on Computer Vision, pages 308#313, May 1993.

....modal framework is that objects must be described in term of the modes of some prototype shape. Such a procedure implicitly imposes an a priori parameterization upon the sensor data. It is thus more suitable for modeling than for tracking purposes. To address this problem, Sclaroff and Pentland[55] recently developed a new method that computes the object s vibration modes directly from the image data. Nastar and Ayache [45] have attempted to unify the work of Terzopoulos et al. and Pentland et al. They followed similar physics principles and developed elastic models for non rigid motion ....

S. Sclaroff and A. Pentland. A modal framework for correspondence and description. Proc. of 12th ICPR, pages 308--313, 1993.

....we have used flexible models templates that deform to fit the image evidence. Given the inherent shape variability of anatomical objects, this approach is well suited to segmentation of medical imagery. Many kinds of flexible template have been described in the literature, for example [7 14]. The major drawback with many of these is that they lack specificity: they can produce illegal examples of the desired objects because they lack global shape constraints. To apply such constraints, the model needs to incorporate a priori knowledge of the objects being segmented. We have used ....

S. Sclaroff, A. Pentland, A modal framework for correspondence and description. In 4 th International Conference on Computer Vision, 308--313, Berlin (May 1993).

....the correspondence problem in image sequence analysis. Objects in the world can be nonrigid, and an object s appearance can deform as the viewing geometry changes. Consequently, much work has also been done that addresses the problem of correspondence and description by using deformable models[10 11,17 19]. Scott and Longuet Higgins [10] developed an algorithm to determine the possible correspondences of 2D point features across a pair of images without any other information (in particular, they had no information about the poses of the cameras) They first incorporated a proximity matrix ....

....best matches. Then they used the eigenvectors of this matrix to determine correspondences between two sets of feature points. Shapiro and Brady [11] also proposed an eigenvector approach to determining point feature correspondence based on a modal shape description. Recently, Sclaroff and Pentland [19] described a modal framework for correspondence and description. In this paper, we first investigate the problem of determining image point correspondences given the poses of two images while simultaneously computing the corresponding 3D points. Here, camera pose consists of an orientation R l ....

S. Sclaroff and A. Pentland, "A modal framework for correspondence and description ", Proc. of IEEE, pp. 308-313, 1993.

....avoid the call to costly eigenvectorextraction routines ; moreover, they allow the total number of modes P Theta P 0 to be easily adjusted. 4. 2 Strain energy as a similarity measure The strain energy describing the amount of energy needed to deform a surface into another is by definition [1, 23] : E strain = 1 2 U T KU This expression is much simpler when developed in the modal (or vibration) basis of the system : E strain = 1 2 N X p=1 N 0 X p 0 =1 2 (p; p 0 ) u 2 (p; p 0 ) Therefore, the low order strain energy corresponding to P Theta P 0 low order modes is ....

S. Sclaroff and A. Pentland. A modal framework for correspondence and description. In Proceedings of the Fourth International Conference on Computer Vision (ICCV '93), Berlin, May 1993.

No context found.

Sclaroff, S., and Pentland, A., A modal framework for correspondence and description. In Proc. of IEEE, pp. 308-313, 1993.

No context found.

S. Sclaroff and A. Pentland. A modal framework for correspondence and description. In Proc. Fourth Internation Conference on Computer Vision, May 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