Advanced

Aligning shapes by minimising the description length

Ericsson, Anders LU and Karlsson, Johan LU (2005) In Lecture Notes in Computer Science 3540. p.709-718
Abstract
When building shape models, it is first necessary to filter out the similarity transformations from the original configurations. This is normally done using Procrustes analysis, that is minimising the sum of squared distances between the corresponding landmarks under similarity transformations. In this article we propose to align shapes using the Minimum Description Length (MDL) criterion. Previously MDL has been used to locate correspondences. We show that the Procrustes alignment with respect to rotation is not optimal. The MDL based algorithm is compared with Procrustes on a number of data sets. It is concluded that there is improvement in generalisation when using Minimum Description Length. With a synthetic example it is shown that... (More)
When building shape models, it is first necessary to filter out the similarity transformations from the original configurations. This is normally done using Procrustes analysis, that is minimising the sum of squared distances between the corresponding landmarks under similarity transformations. In this article we propose to align shapes using the Minimum Description Length (MDL) criterion. Previously MDL has been used to locate correspondences. We show that the Procrustes alignment with respect to rotation is not optimal. The MDL based algorithm is compared with Procrustes on a number of data sets. It is concluded that there is improvement in generalisation when using Minimum Description Length. With a synthetic example it is shown that the Procrustes alignment can fail significantly where the proposed method does not. The Description Length is minimised using Gauss-Newton. In order to do this the derivative of the description length with respect to rotation is derived. (Less)
Please use this url to cite or link to this publication:
author
organization
publishing date
type
Contribution to journal
publication status
published
subject
in
Lecture Notes in Computer Science
volume
3540
pages
709 - 718
publisher
Springer
external identifiers
  • wos:000230372500072
  • scopus:26444484303
ISSN
1611-3349
DOI
10.1007/b137285
language
English
LU publication?
yes
id
b1f39688-9d9a-475d-afe1-2c6774d0f08c (old id 231563)
date added to LUP
2007-08-02 16:06:25
date last changed
2017-01-01 05:09:51
@article{b1f39688-9d9a-475d-afe1-2c6774d0f08c,
  abstract     = {When building shape models, it is first necessary to filter out the similarity transformations from the original configurations. This is normally done using Procrustes analysis, that is minimising the sum of squared distances between the corresponding landmarks under similarity transformations. In this article we propose to align shapes using the Minimum Description Length (MDL) criterion. Previously MDL has been used to locate correspondences. We show that the Procrustes alignment with respect to rotation is not optimal. The MDL based algorithm is compared with Procrustes on a number of data sets. It is concluded that there is improvement in generalisation when using Minimum Description Length. With a synthetic example it is shown that the Procrustes alignment can fail significantly where the proposed method does not. The Description Length is minimised using Gauss-Newton. In order to do this the derivative of the description length with respect to rotation is derived.},
  author       = {Ericsson, Anders and Karlsson, Johan},
  issn         = {1611-3349},
  language     = {eng},
  pages        = {709--718},
  publisher    = {Springer},
  series       = {Lecture Notes in Computer Science},
  title        = {Aligning shapes by minimising the description length},
  url          = {http://dx.doi.org/10.1007/b137285},
  volume       = {3540},
  year         = {2005},
}