Affine-invariant geometric shape priors for region-based active contours.

Area of Science:

• Computer Vision
• Image Processing
• Computational Geometry

Background:

• Active contours are widely used for image segmentation.
• Constraining active contour evolution with shape priors is crucial for accurate segmentation.
• Existing methods may lack robustness to pose and affine deformations.

Purpose of the Study:

• To develop a new method for constraining region-based active contour evolution.
• To incorporate a shape prior based on Legendre moments for improved segmentation.
• To achieve intrinsic invariance to pose and affine deformations in the shape model.

Main Methods:

• Defining a shape prior as the distance between shape descriptors derived from Legendre moments of the characteristic function.
• Minimizing this shape prior to derive a geometric flow.
• Applying the geometric flow in a two-class image segmentation context.

Main Results:

• The proposed method effectively constrains active contour evolution.
• The Legendre moments-based shape prior leads to a beneficial geometric flow.
• The shape model demonstrates intrinsic invariance to pose and affine deformations.

Conclusions:

• The new method offers an effective way to constrain active contour evolution using shape priors.
• Legendre moments provide a robust basis for shape descriptors in segmentation.
• The developed shape model enhances segmentation accuracy and robustness.