Presentations and Results:

In the following you can find Martin Drauschke's results of the second period, which have been presented to the bundel project partners at the regular meetings.

The main topics of our current research are:

Further down on this page, there are presented the old results from the first period (by Axel Gotthardt).



Stability of Regions in Scale Space:

Presentation at bundel meeting on 20th July 2006: Automatic Tracing of Stable Regions (in German), showing the following movies (stable regions are marked red):

Presentation at bundel meeting on 18th May 2006: Investigation of Stable Regions in Scale Space for Building Detection (in German)


Dormer Detection in Aerial Images

Results may appear soon.



Results by Axel Gotthardt:

In the following you can find some of our results .

The main topics of our research within the first period have been:




Automated Building Detection:

We realised a method for detecting building roofs in aerial images. The detection model is a Bayesian net that aggregates cliques of image regions which may cover a complex object. Observable attributes of the regions are derived from a rich symbolic image description containing points, lines and regions as basic features including their relations (Fuchs/Förstner 1995), Images 1+2). The model captures the dependency of region aggregates together with the neighbouring features with respect to observability due to occlusions and to perspective deformations. Up to now, the model captures cliques of the region adjacency graph (RAG) with one, two and three regions which is sufficient for detecting polyhedral objects. Cliques are classified using MAP estimation. The model allows to detect and locate multiple appearances of object classes. The joint distribution of the Bayesian net is determined in a supervised learning step based on images with annotated regions. The method is realized and demonstrated for the detection of building roofs in aerial images (images 3+4).

Further details on our method for roof detection can be found in the technical report (TB-ipb-05_1.pdf).



1: RAG
2: Building a Graphical Model
3: Extracted Features
4: Detected Roof Regions



Realisation of a Tool Chain:





Improvement of the Extraction of Regions


For investigating the effect of scale onto building detection we extend the used feature extraction scheme especially in textured areas. We analyse the usefullness of color spaces and the change of texture in scale space for the extraction of regions.

The following results of segmentation have been achieved using color histograms over several scales:


Source Image
Detected Roof region
Detected Vegetation
Detected Street regions