Browse by author
Lookup NU author(s): Stephen Purves, Emeritus Professor Satnam Dlay
Full text for this publication is not currently held within this repository. Alternative links are provided below where available.
The ability to describe and accurately measure shape in digital images is a useful tool in a number of applications. The performance of methods such as scene reconstruction, object segmentation and object recognition and classification can be facilitated by low level shape descriptors. However, curved image boundaries are seriously degraded by the imaging process itself and accurate curvature measurements are difficult to recover. In this paper, a novel approach to discrete curvature estimation is presented that makes use of the varying responses of spatial filters to curved edge stimuli. A limited set of basis filters are used to detect edge features and their apparent orientations. A second set of elliptical filters are then applied to provide an estimate of local curvature. The resulting curvature measurement is invariant to the contrast, relative position and orientation of the feature and is suitable for use with a curve/contour model.
Author(s): Purves SJ, Dlay SS
Publication type: Article
Publication status: Published
Journal: Advances in Physics, Electronics and Signal Processing Applications
Year: 2000
Pages: 246-251
Print publication date: 01/01/2000
ISSN (electronic):
URL: http://www.wseas.us/e-library/conferences/athens2000/Papers2000/427.pdf
