A discrete dynamic contour model

Area of Science:

• Computer Vision
• Image Processing
• Computational Geometry

Background:

• Active contour models are widely used for image segmentation.
• Existing models often suffer from undesirable deformation effects such as shrinking and vertex clustering.
• A robust and reproducible contour definition method is needed.

Purpose of the Study:

• To develop a discrete dynamic model for defining contours in 2-D images.
• To address limitations of existing active contour models, specifically shrinking and vertex clustering.
• To achieve reproducible approximations of desired contours through energy minimization.

Main Methods:

• A discrete dynamic model composed of connected vertices is proposed.
• The model is initialized with minimal user interaction.
• An energy minimizing process modifies the contour, with internal energy based on curvature and external energy on image features.

Main Results:

• The model successfully avoids shrinking and vertex clustering.
• The deformation process converges to a local minimum of the energy function.
• The method yields reproducible contour approximations for both synthetic and clinical images.

Conclusions:

• The developed discrete dynamic active contour model offers an effective solution for image segmentation.
• It overcomes common limitations of existing methods, providing stable and accurate contour definition.
• The model's reproducibility and robustness make it suitable for various imaging applications.