Survey of Distance Transforms

About

Code

Lenna

Inscribed Circle

About

This is the companion website to the paper

"2D Euclidean distance transforms: a comparative survey", ACM Computing Surveys, Vol 40, Issue 1, Feb 2008. (pdf | bib | errata)

A more complete set of results is posted here, together with source code for the algorithms and the performance benchmark. Further results and errata may also be posted in the future. In the results sections of this page you will find test images and performance plots, along with distance histograms characterizing the distance maps of the test images.

Feel free to contact us if you need additional information or help with the source code.

Authors

Ricardo Fabbri, Luciano da F. Costa, Julio C. Torelli, and Odemir M. Bruno.

Press

"Olhos de Máquina", Agencia FAPESP, 07/04/2008

Code Download

Animal C library
- A simple portable library containing the C implementation of fast exact 2D Euclidean distance transforms used in the Survey.
SIP Toolbox
- Toolbox for Scilab (free software similar to Matlab) containing easy-to-use scripted interface to the distance transform algorithms used in the survey.
TeDi Benchmark Suite (GNU/Linux)
- A collection of Shell and Scilab scripts controlling the tests reported in the survey, as well as C code to generate some of the datasets. This package enables the user to run tests in their own computer and report the results to be published in this website. The user can generate test datasets, measure algorithm performance, add their own algorithm, view performance plots, add new tests, and so on.
- Subversion Repository Access:
  svn co https://distance.svn.sourceforge.net/svnroot/distance/trunk/tedi
- The synthetic dataset used in the paper: (download ~70MB)
- Sourceforge record for the project
VxL C++ Computer Vision Libraries
- The fastest Euclidean distance transform algorithms were ported to this C++ library by the author of the survey.
- The code is in vil/core/algo/vil_exact_distance_transform.{cxx,h}.
  For example usage, see vil/core/algo/tests/.
- A 3D implementation of Saito is also available at: contrib/brl/bbas/bil/algo/bil_edt.{cxx,h}