Related Experiment Videos
A shape representation for computer vision based on differential topology
Bio Systems
|January 1, 1995
Summary
We introduce a novel shape representation for computer vision using graph structures derived from level sets. This stable method simplifies object recognition and enables symbolic smoothing for coarse-to-fine strategies.
Area of Science:
- Computer Vision
- Computational Geometry
- Differential Topology
Background:
- Object recognition relies on effective shape representation.
- Existing methods may lack stability or simplicity under deformation.
Purpose of the Study:
- To present a new shape representation for computer vision.
- To leverage differential topology for robust shape analysis.
- To enable efficient object recognition and deformation analysis.
Main Methods:
- Utilizing graph structures derived from level sets.
- Applying principles from differential topology and singularity theory.
- Developing a representation stable under deformation.
Main Results:
- The proposed representation is both stable and exhibits simple changes under deformation.
- Introduces 'symbolic smoothing' for representation domain smoothing.
- Demonstrates applicability to 2D and 3D objects, including silhouettes.
Conclusions:
- The novel shape representation offers advantages in stability and simplicity.
- Symbolic smoothing facilitates coarse-to-fine recognition strategies.
- The representation is versatile for various object types and dimensions.