Anti-Impulse-Noise Edge Detection via Anisotropic Morphological Directional Derivatives.

Area of Science:

• Computer Vision
• Image Processing
• Signal Processing

Background:

• Traditional edge detection methods degrade significantly with impulse noise.
• Morphological operators like median filters are effective against impulse noise.

Purpose of the Study:

• To develop an impulse noise-robust edge detection method using anisotropic morphological directional derivatives (AMDDs).
• To enhance edge detection performance in noisy image conditions.

Main Methods:

• Proposed anisotropic morphological directional derivatives (AMDDs) by combining biwindow configuration with weighted median filters.
• Derived AMDD spatial response and directional representation for ideal step edges.
• Utilized spatial and directional matched filters to extract edge strength maps (ESM) and edge direction maps (EDM).
• Integrated ESM and EDM into differential-based edge detection for an anti-impulse-noise detector.

Main Results:

• The proposed AMDD-based edge detector demonstrates competitive performance in noise-free and Gaussian noise scenarios.
• Achieved superior performance compared to state-of-the-art detectors specifically in impulse noise conditions.
• Analysis of biwindow characteristics and edge resolution provided insights into detector behavior.

Conclusions:

• The AMDD-based edge detector offers a robust solution for edge detection in the presence of impulse noise.
• The method effectively preserves edge information while mitigating noise interference.
• This approach advances the field of noise-resilient image processing.