CVXB

CVXB performs artifact-aware Chan-Vese segmentation with Retinex-inspired bias correction to separate piecewise-constant structural components from smooth bias fields and isolate intensity outliers for robust analysis of images affected by artifacts and illumination biases.

Key Features:

• Dynamic Artifact Class: Introduces a dynamic artifact class within a single energy functional to identify and isolate intensity outliers that would otherwise distort segmentation.
• Retinex-inspired Decomposition: Decomposes images into a piecewise-constant structural part and a smooth bias part so segmentation terms operate on the structural component only.
• Phase-field Parameterization: Parameterizes segmentation with a phase-field approach to represent region boundaries continuously.
• Threshold-dynamics Minimization: Employs threshold dynamics for efficient minimization of the energy functional.
• Chan-Vese Integration: Integrates Chan-Vese (CV) segmentation terms applied solely to the structural component while excluding identified artifact regions.
• Rapid Convergence: Typically converges within 10–50 iterations, producing meaningful results in fractions of a second on standard computing equipment.

Scientific Applications:

• Medical Imaging: Applied to lesion detection and bias field correction in modalities such as magnetic resonance imaging (MRI).
• Artifact-Compromised Images: Segmentation and analysis of images damaged by artifacts or intensity outliers that interfere with contrast.
• Diverse Imaging Modalities: Demonstrated effectiveness across a variety of sample images from different imaging modalities.

Methodology:

Retinex-inspired decomposition into piecewise-constant structural and smooth bias components; incorporation of a dynamic artifact class in a single energy functional; phase-field parameterization with threshold-dynamics minimization of Chan-Vese terms applied only to the structural component; typical convergence in 10–50 iterations.