DTU > IMM > DSP > THOR >>> HBP > Projects > Geo


Series of icons with a surface fitting a shape

“Deformable surface project 1997”


Icon with surface fitting a
	    head Shell and Head [1.6MB MPEG1]
Head [1.1MB MPEG1]
Shell [0.6MB MPEG1]
Shell [1.1MB Animated GIF]

In many areas of science a general visualisation problem exists, namely that the fundamental object is a three dimensional structure embedded in a rectangular volume. One example could be data from medical brain scanners. Normally it is qualitatively easy for the human to see and identify the structure from, e.g., images of the individual slices through the volume. It is, however, desirable to have automatic procedures, working in 3D, to track the boundary of the object and to generate a closed set of triangles (polygons). These closed surfaces can be used to visualise the object in 3D, or to perform quantitative measurements within the structure.

In the paper [1] a generic way to track the surface of 3D objects from volumes is presented. Our strategy is to shape a closed surface of connected triangles to match the boundary of the object. There is no easy way to solve this non-linear inversion problem, so the adaptive strategy we suggest invokes minimisation of an energy function containing a set of individually weighted terms reflecting a set of desired goals.
  1. Mattias Ohlsson, Peter Toft, Lars Kai Hansen and Finn Nielsen
    Active Surface Models for Brain Imaging
    Presented at the Interdisciplinary Inversion Workshop 5, Århus University, September 24 1997.
    Keywords: Mesh fitting, Object tracking, 3D modelling
    [ Expanded abstract (ps.gz) | Poster (ps.gz) | paper (ps.gz) ]
Mattias Ohlsson, Peter Toft, Finn Årup Nielsen; 1997 October.


Morphing icon


[VRML2, 300kb] Brain Morphing. Watch the optimization of a model of a head: A 'finite difference model' is constructed, and given a gradient image and an initial guess for the surface the model is optimized with gradient descent steps. This specific model is interpolating between the first initial guess and the 1000th optimization step.

A bigger model [VRML2, 2.5Mb]has every 25th step between the initial guess and the 1000th step.

The model contains a small user interface: The green arrow should start the animation. The red square will stop it. After you have stopped the animation you can manually move through the steps with the slider1. Note that any smooth interpolation between the single steps in the optimization is done by VRML, - not by the optimization.

More VRML files are available from the VRML page.
  1. Slider by Rich Gossweiler VRML-employee at SGI, http://reality.sgi.com/rcg/
Finn Årup Nielsen; 1997.


Icon with surface fitting a shape


The developed software is now available for free (protected by the GNU public license). Press here to download the software. The software is distributed as is. The main program called geo is written in C using standard Unix tools. The programs should compile on most Unix systems. Linux and SGI has been tested. Unpack the geo.tar.gz using

gunzip geo.tar.gz
tar xvf geo.tar

This makes a directory named geo containing the code. Look into the GNU file to see the copy protection policy, and the file named README contains some guide to what is found where. Each of the C-functions in the geo/sourcedir directory contains descriptions of the functions. We encourage you to read it. The main control function main is found in geo/sourcedir/geo.c - trace down the code from here to understand the core of the programs.
Mattias Olhsson; Peter Toft; 1997.


icon with brain HBP THOR Center for Neuroinformatics, Human Brain Project Repository (This server)
THOR THOR Center for Neuroinformatics
DSP Digital Signal Processing
SIP Section for Signal/Image Processing
IMM Informatics and Mathematical Modelling
DTU Technical University of Denmark


© IMM DTU, 1997, 2001, 2002, 2007
$Id: geo.html,v 1.4 2007/08/01 11:01:42 fn Exp $