On Two Stages for Refined Building Roof Extraction in an Aerial Image

David Tien, Xiao Yi, Tony Hope

Research output: Book chapter/Published conference paperConference paperpeer-review

2 Citations (Scopus)
64 Downloads (Pure)


Robust segmentation of building roofs is desired in urban and regional planning and asset management. In this paper, we present a two-stage method for building roof region extracting using the cues of colour and edges. At the first stage, three layers are generated. The first and second layer locate initial buildingroof regions by the colour based segmentation processes at two different threshold values to obtain an under-segmentation and an over-segmentation result, respectively. The third layer divides the whole area of the aerial image into pieces by an edge detector. At the second stage, we refine the regions of located building roofs using a combination of regions found in the three layers. Experimental results show that the proposed approach yields intuitively correct as well as accurate extraction of the building roof regions in urban aerial images.
Original languageEnglish
Title of host publication4th ICITA 2007
EditorsShi Guanfan
Place of PublicationSydney, Australia
PublisherMacquarie Scientific Press
Number of pages4
ISBN (Electronic)9780980326703
Publication statusPublished - 2007
EventInternational Conference on Information Technology and Applications (ICITA) - Harbin, China, China
Duration: 15 Jan 200718 Jan 2007


ConferenceInternational Conference on Information Technology and Applications (ICITA)

Fingerprint Dive into the research topics of 'On Two Stages for Refined Building Roof Extraction in an Aerial Image'. Together they form a unique fingerprint.

Cite this