1. Summary
  2. Files
  3. Support
  4. Report Spam
  5. Create account
  6. Log in


From composight

(Redirected from Main Page)
Jump to: navigation, search

Composight is a C++ toolkit for 3D vision in the domain of composite materials science. It provides small, problem specific applications for viewing, filtering, segmentation of volumetric data such as micro CT scans. Our goal is to make some useful Algorithms from the Insight Toolkit available for material researchers with no in-depth programming education. In order to achieve this, ITK components are assembled to small commandline tools, each aiming at one specific topic. All tools are scriptable by either a sequence of simple commands, like "Load example.mha", or ECMAScript, aka "JavaScript". The Segmentation Program additionally provides a GUI version, which can be used to evaluate some filter parameters on a small subsection of the image (ROI) interactively and to store some good paramters to disk. Parameters can be reloaded in the commandline counterpart in order to process larger Images.


Work in Progress

Detailed build instructions for our main target platform Visual Studio 2008 on Windows x64. So far it seems to cause less problems if you just install all development packages of ITK, VTK and Qt on a UNIX machine. Then check out https://composight.svn.sourceforge.net/svnroot/composight/trunk/, run CMake and make to play around with the Segment application.

We encountered some problems with ITK on Windows x64 which we are currently about to fix in a vendor branch of ITK in the composight repository. This will by Insight developers as soon as version 4.xx is out. There is no attempt to fix it in ITK 3.xx. However you can try an adapted version of ITK 3.20. Using it for Visual Studio builds, larger Images can be processed in RAM but another issue with large file support in the MS STL restricts the filesize of stored Images to 2 GB.

So for now we tested and recommend using Ubuntu 10.10 x64. Most other UNIXes that provide Qt 4.x, ITK 3.xx and VTK 5.x should work as well.

Currently we prepare "Orient" for publication on sourceforge. It is designed to measure local fibre orientations on segmented images, to visualize that direction and write it out to an oriantation histogram. Two methods implemented: A slow but very noise tolerant algorithm that computes the strongest of a set of nonisotropic gaussian filterresponses. The second method makes use of characteristic Hessian behaviour within fibers.

Python scripting will be available by using PythonQt.

The 3rd program to be published is for computing statistics on separate connected components of a segmentation. In case of fibers or pores this means that for each single fibre or pore, geometrical data like fit ellipses or volume and surface curvature statistics can be computed.


If you are using composight for your publications, please give it credit by citing the publications by the Institute of Materials Science and Engineering (IWK I as part of the KIT) that led to composight.



Project funding is provided by the Institute for Applied Materials which is part of the Karlsruhe Institute of Technology


How to contribute to this Mediawiki

Personal tools