A Distributed and Collective Approach for Curved Object-Based Range Image Segmentation

Abstract : In this paper, we use multi-agent paradigm in order to propose a new method of image segmentation. The images considered in this work are the range images which can contain at once polyhedral and curved objects. The proposed method uses a multi-agent approach where agents align the region borders to the surrounding surfaces which make emerging a collective segmentation of the image. The agents move on the image and when they arrive on the pixels of a region border they align these pixels to their respective surfaces. The resulting competitive alignment allows at once the emergence of the image edges and the disappearance of the noise regions. The test results obtained with real images show a good potential of the new method for accurate image segmentation.
Complete list of metadatas

https://hal-lirmm.ccsd.cnrs.fr/lirmm-00520949
Contributor : Fabien Michel <>
Submitted on : Friday, September 24, 2010 - 3:42:59 PM
Last modification on : Thursday, October 31, 2019 - 9:22:02 PM

Links full text

Identifiers

Citation

Smaine Mazouzi, Zahia Guessoum, Fabien Michel. A Distributed and Collective Approach for Curved Object-Based Range Image Segmentation. 14th Iberoamerican Conference on Pattern Recognition, CIARP 2009, Nov 2009, Guadalajara, Mexico. pp.201-208, ⟨10.1007/978-3-642-10268-4_24⟩. ⟨lirmm-00520949⟩

Share

Metrics

Record views

216