Non-Euclidean basis function based level set segmentation with statistical shape prior.

Area of Science:

• Computer Vision
• Medical Image Analysis
• Computational Geometry

Background:

• Level set methods are widely used for image segmentation.
• Statistical shape models (SSMs) provide powerful shape priors.
• Integrating SSMs with level sets can improve segmentation accuracy and robustness.

Purpose of the Study:

• To develop a new image segmentation framework combining SSMs and level sets.
• To represent level sets using non-Euclidean radial basis functions (RBFs).
• To eliminate the need for reinitialization in level set evolution.

Main Methods:

• The proposed framework represents level sets as a linear combination of non-Euclidean RBFs.
• Shape priors are encoded as probabilistic maps from training data.
• The RBF representation allows level set evolution to be described by ordinary differential equations.

Main Results:

• The method achieves accurate and topologically flexible image segmentation.
• Incorporating image information into RBFs enhances segmentation performance.
• Experimental results demonstrate the advantages over traditional level set methods.

Conclusions:

• The new framework offers an efficient and effective approach to image segmentation.
• The use of non-Euclidean RBFs provides significant improvements in accuracy and flexibility.
• This method represents a valuable advancement in statistical shape model-enhanced level set segmentation.