Level set model with local fitting operation of median filter

Area of Science:

• Computer Vision
• Image Processing
• Medical Imaging

Background:

• Traditional local binary fitting operations, such as Gaussian kernel function weighted summation (GKFWS), are fundamental in image segmentation.
• Existing methods may face challenges with noise and intensity inhomogeneity, impacting segmentation accuracy.
• Level set models offer a powerful framework for image segmentation but require refinement for robustness.

Purpose of the Study:

• To develop a novel level set model that enhances image segmentation performance.
• To address limitations of traditional methods concerning noise and intensity variations.
• To improve the computational efficiency and robustness of image segmentation techniques.

Main Methods:

• The proposed model replaces the traditional Gaussian kernel function weighted summation (GKFWS) with a median filter in the local fitting procedure.
• Incorporation of the edge stopping function from the Geodesic Active Contour (GAC) model to integrate edge information.
• Utilized a level set framework for image segmentation.

Main Results:

• Experimental results on synthetic and real images demonstrate the model's effectiveness.
• The novel model shows promising performance in terms of computational cost.
• The method exhibits robustness to noise and effective segmentation of images with intensity inhomogeneity.

Conclusions:

• The developed level set model offers a significant improvement over traditional methods.
• The integration of median filtering and edge stopping functions enhances segmentation accuracy and robustness.
• This approach provides a valuable tool for various image segmentation applications.