Complementary texture and intensity gradient estimation and fusion for watershed segmentation

Corcoran, Padraig and Winstanley, Adam C. and Mooney, Peter (2011) Complementary texture and intensity gradient estimation and fusion for watershed segmentation. Machine Vision and Applications, 22. pp. 1027-1045. ISSN 0932-8092

Download (2MB) | Preview

Share your research

Twitter Facebook LinkedIn GooglePlus Email more...

Add this article to your Mendeley library


In this paper, we identify two current challenges associated with watershed segmentation algorithms which attempt to fuse the visual cues of texture and intensity. The first challenge is that most existing techniques use a competing gradient set which does not allow boundaries to be defined in terms of both visual cues. The second challenge is that these techniques fail to account for the spatial uncertainty inherent in texture gradients. We present a watershed segmentation algorithm which provides a suitable solution to both these challenges and minimises the spatial uncertainty in boundary localisation. This is achieved by a novel fusion algorithm which uses morphological dilation to integrate intensity and texture gradients.Aquantitative and qualitative evaluation of results is provided demonstrating that our algorithm outperforms three existing watershed algorithms.

Item Type: Article
Additional Information: The definitive version of this article is available at DOI: 10.1007/s00138-010-0310-z © Springer-Verlag 2010
Keywords: Feature fusion; Spatial uncertainty; Texture; Watershed segmentation;
Academic Unit: Faculty of Science and Engineering > Computer Science
Faculty of Science and Engineering > Research Institutes > National Centre for Geocomputation, NCG
Item ID: 5830
Identification Number: 10.1007/s00138-010-0310-z
Depositing User: Peter Mooney
Date Deposited: 16 Feb 2015 17:24
Journal or Publication Title: Machine Vision and Applications
Publisher: Springer Verlag
Refereed: Yes

Repository Staff Only(login required)

View Item Item control page

Document Downloads

More statistics for this item...