Title: An integrated approach to the segmentation and recognition of objects using level set and thin plate spline method

Authors: Xun Wang, Feng Gao, Zhigang Peng, William G. Wee

Addresses: Aureon Biosciences Corporation, 28 Wells Avenue, Yonkers, NY 10701, USA. ' Department of Mathematics and Computer Science, Claflin University, Orangeberg, SC 29118, USA. ' Department of ECECS, University of Cincinnati, Cincinnati, OH 45220, USA. ' Department of ECECS, University of Cincinnati, Cincinnati, OH 45220, USA

Abstract: A novel approach for object segmentation and recognition is presented. The aim of this approach is to select a proper shape model from a model set to guide object segmentation. The process of model selection, which is based on the shape similarity between the target object and shape models, is then used for object recognitions. The integrated process of object segmentation and recognition is formulated as a constrained contour energy minimisation problem. The solution derived from this formulation produces an integrated searching process consisting of two iteratively alternating procedures of contour evolution and shape matching. The process stops at a final contour together with a shape distance measure to an object model for recognition. Successful illustrative results on both segmentation and recognition are reported.

Keywords: level sets; curve evolution; variational methods; shape matching; object segmentation; object recognition; thin plate splines; shape models; contour evolution; object modelling.

DOI: 10.1504/IJISTA.2007.014260

International Journal of Intelligent Systems Technologies and Applications, 2007 Vol.3 No.3/4, pp.211 - 225

Published online: 28 Jun 2007 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article