A Variational Level Set Approach Based on Local Entropy for Image Segmentation and Bias Field Correction

Area of Science:

• Medical Image Analysis
• Computer Vision

Background:

• Image segmentation is challenging due to intensity inhomogeneity (bias field).
• Existing methods struggle with accurate bias field estimation and segmentation simultaneously.

Purpose of the Study:

• To present a novel region-based approach for simultaneous image segmentation and bias field estimation.
• To improve accuracy in both segmentation and bias field correction.

Main Methods:

• A local Gaussian distribution fitting (LGDF) energy function weighted by local entropy is proposed.
• The method incorporates bias field prior information for enhanced accuracy.
• Level set regularization is used for energy function minimization.

Main Results:

• The proposed method effectively segments images and estimates the bias field concurrently.
• Experiments show superior performance compared to state-of-the-art approaches across various image modalities.
• Accurate bias field estimation is achieved due to the model's design.

Conclusions:

• The novel region-based approach offers a robust solution for image segmentation with simultaneous bias field estimation.
• This method demonstrates significant improvements in accuracy and performance for medical image analysis.
• The technique is effective across diverse imaging modalities.