eTRIMS - E-Training for Interpreting Images of Man-Made Scenes
 
eTRIMS
     
project results:

Fully bottom up blob detection in building facades

Mohammad Jahangiri, Maria Petrou
Electrical and Electronics Department, Communication and Signal Processing Group
Imperial College of London
June, 2008

A fully bottom up algorithm for extracting blobs in building facades is proposed. The proposed extraction mechanism exploits the fact that the most interesting parts of an image are those that contain some variation i.e. local texture. A scheme independent of scale and colour is proposed, based on the use of Gaussian kernels and mathematical morphology. Prototype discovery, hypothesis testing and verification are used to identify any missed blobs when repeated patterns are present in the facade. A novel metric is proposed for assessing the similarity of sub-patterns. We evaluate the proposed technique using 300 annotated building facades which were collected from 4 different European cities. The proposed algorithm detected 80% of the 3589 hand segmented components.

Bounding boxes of the preliminary detected blobs

Prototype Discovery

Bounding boxes of the final result

References
under review