Corner Detection Using Second-Order Generalized Gaussian Directional Derivative Representations.

Area of Science:

• Computer Vision
• Image Processing
• Pattern Recognition

Background:

• Corner detection is vital for image analysis tasks like object recognition.
• Current algorithms struggle to differentiate between edges and corners, leading to detection errors.

Purpose of the Study:

• To evaluate generalized Gaussian directional derivative filters for noise suppression.
• To develop a new corner detection method addressing limitations of existing approaches.

Main Methods:

• Investigated second-order generalized Gaussian directional derivative representations for various corner types.
• Utilized properties of edges and corners to propose a novel detection algorithm.
• Evaluated detector performance based on accuracy, repeatability, and region repeatability.

Main Results:

• The proposed detector accurately distinguishes corners from edges, reducing false detections.
• Demonstrated effective Gaussian noise suppression using generalized Gaussian directional derivative filters.
• Achieved superior performance over nine state-of-the-art methods under various transformations and degradations.

Conclusions:

• The novel corner detection method offers improved accuracy and robustness.
• Generalized Gaussian directional derivative filters show promise for noise suppression in image analysis.
• This work advances corner detection for critical image understanding applications.