Related Experiment Video
Updated: Jun 24, 2026
14:08
Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
Published on: April 13, 2013
Subpixel edge refinement using deformable models
Frédéric Bouchara1, Sofiane Ramdani
1UMR CNRS 6168, Laboratoire des Sciences de l'Information et des Systèmes, Université du Sud Toulon-Var, 83957 La Garde Cedex, France. bouchara@univ-tln.fr
Summary
This study introduces a novel subpixel edge refinement technique using deformable models. The method enhances image edge accuracy by integrating gradient orientation and position information for superior detail.
Area of Science:
- Computer Vision and Image Processing
- Computational Geometry
Background:
- Accurate edge detection is crucial for image analysis.
- Existing methods often struggle with subpixel precision.
Purpose of the Study:
- To develop a robust subpixel edge refinement method.
- To improve the accuracy of edge localization in digital images.
Main Methods:
- Utilizing deformable models constrained by gradient orientation and position information.
- Formulating the problem within a statistical maximum a posteriori (MAP) estimation framework.
- Proposing both a parametric (B-spline) and a sampled deformable model.
Main Results:
- Demonstrated effectiveness on both synthetic and natural image datasets.
- Achieved significant improvements in subpixel edge localization accuracy.
- Validated the proposed algorithms' adequacy and performance.
Conclusions:
- The proposed deformable model approach effectively refines subpixel edges.
- Integration of orientation and position information enhances edge detection precision.
- The statistical framework provides a solid foundation for edge refinement algorithms.
