Fast Parabola Detection Using Estimation of Distribution Algorithms.

Area of Science:

• Computer Vision
• Medical Image Analysis
• Computational Intelligence

Background:

• Parabolic shape detection is crucial in various image analysis tasks.
• Existing methods like Hough transform and RANSAC have limitations in efficiency.
• Developing faster and accurate parabolic shape detection algorithms is needed.

Purpose of the Study:

• To introduce a novel method for parabolic shape detection using Estimation of Distribution Algorithms (EDAs).
• To evaluate the computational efficiency and accuracy of the proposed EDA-based method.
• To compare the performance against generalized Hough transform and RANSAC.

Main Methods:

• A virtual parabola is generated using three random boundary pixels.
• The parabola's constants are calculated from these pixels.
• The Hadamard product serves as the fitness function for matching the virtual parabola to image features.

Main Results:

• The proposed EDA-based method achieved significant speed improvements: 93.61% on synthetic images and 89% on medical images.
• The method demonstrated superior computational time compared to generalized Hough transform and RANSAC.
• High suitability for diverse medical imaging applications was observed.

Conclusions:

• The EDA-based method offers a highly efficient and accurate solution for parabolic shape detection.
• This approach presents a valuable tool for medical image analysis, including retinal fundus and plantar arch images.
• The method's speed and accuracy make it a promising alternative to traditional techniques.