Guide filter-based gradient vector flow module for infrared image segmentation.

Area of Science:

• Computer Vision
• Image Processing
• Artificial Intelligence

Background:

• Infrared image segmentation is challenging due to noise, low contrast, and weak edges.
• Active contour models like Gradient Vector Flow (GVF) struggle with noisy infrared images.
• Existing GVF models are sensitive to noise and parameter adaptability, limiting their effectiveness.

Purpose of the Study:

• To propose a novel guide filter-based gradient vector flow (GFGVF) module for infrared image segmentation.
• To overcome the limitations of traditional GVF models in noisy infrared image segmentation.
• To improve segmentation accuracy, noise robustness, and boundary leakage control.

Main Methods:

• Utilized a guide filter to create a novel edge map, effectively capturing image edges while mitigating noise.
• Developed a new weighting function to enhance the capture range and preserve edges in noisy conditions.
• Integrated the guide filter and novel weighting function into the Gradient Vector Flow (GVF) model, creating the GFGVF module.

Main Results:

• The GFGVF model demonstrated a large capture range and concavity convergence.
• Experimental results showed significant improvements in noise robustness compared to traditional methods.
• The proposed method effectively alleviated boundary leakage issues common in infrared image segmentation.

Conclusions:

• The GFGVF module offers a robust solution for infrared image segmentation.
• The integration of guide filter and novel weighting function successfully addresses noise and edge degradation.
• GFGVF provides superior performance in terms of capture range, convergence, noise immunity, and boundary accuracy.