Magazine Article

Finding Known Shapes in an Image by Pruning Parameter Space

TBMG-7024

12/01/2000

Abstract
Content

An improved method of processing two- and three-dimensional image data to locate known shapes called "geometric primitives" involves (1) extraction of edges and other relevant image features and (2) performing a hierarchical search, in a space of parameters of equations that describe the shapes of the features, for those parameters that represent the geometric primitives. This method is inspired by prior object-recognition methods in which parameter spaces are recursively divided and pruned. The most closely related prior methods of this type are based on variations of the Hough transform. Whereas the prior methods have generally offered robustness or computational efficiency but not both, this method offers both, along with other advantages: It enables the efficient and robust extraction of geometric primitives from noisy and incomplete data that include many distracting data, without need for initial estimates of the locations of the geometric primitives.

Meta TagsDetails
Citation
"Finding Known Shapes in an Image by Pruning Parameter Space," Mobility Engineering, December 1, 2000.
Additional Details
Publisher
Published
Dec 1, 2000
Product Code
TBMG-7024
Content Type
Magazine Article
Language
English