Multi-scale-average-filter-assisted level set segmentation model with local region restoration achievements.

Area of Science:

• Computer Vision
• Image Processing
• Computational Imaging

Background:

• Image segmentation is difficult for noisy images with background light and intensity variations.
• Existing methods struggle with inhomogeneity and high-contrast backgrounds.

Purpose of the Study:

• To propose a novel variational method for joint restoration and segmentation of noisy images.
• To address challenges posed by intensity inhomogeneity and high-contrast backgrounds.

Main Methods:

• Combines statistical local region information with a multi-phase segmentation technique.
• Utilizes a fuzzy set framework and alternating direction minimization of multipliers.
• Tested on diverse synthetic and real images with noise and inhomogeneity.

Main Results:

• The proposed model demonstrates superiority over existing two-phase and multi-phase approaches.
• Achieves precise segmentation of images with brightness, diffuse edges, and inhomogeneity.
• Evaluated for precision and robustness on various image types.

Conclusions:

• The novel variational method offers an effective solution for segmenting and restoring challenging noisy images.
• The approach provides accurate segmentation even in the presence of significant image degradation.
• Empirical evaluations confirm the model's efficiency and robustness against state-of-the-art methods.