Speed improvement of B-snake algorithm using dynamic programming optimization

Area of Science:

• Computer Vision
• Image Processing
• Computational Geometry

Background:

• Contour approximation is crucial for image analysis and feature extraction.
• Traditional dynamic programming (DP) methods for contour approximation face challenges with high computational complexity and storage demands.
• The B-snake algorithm offers a flexible framework for curve representation.

Purpose of the Study:

• To develop a novel and computationally efficient approach for contour approximation.
• To reduce the computational complexity and memory requirements compared to standard dynamic programming methods.
• To validate the proposed method's performance on various image types.

Main Methods:

• Integration of the B-snake algorithm with dynamic programming (DP) optimization.
• Development of a new contour point search procedure.
• Experimental evaluation on synthetic, magnetic resonance, and computed tomography medical images.

Main Results:

• Reduced computational complexity from O(N×M(4)) to O(N×M(2)).
• Decreased storage requirements from N×M(3) to N×M memory elements.
• Achieved results equivalent to the standard DP algorithm in contour approximation accuracy.

Conclusions:

• The proposed B-snake and DP integrated approach offers a significant improvement in efficiency for contour approximation.
• This method provides a viable alternative for contour approximation in applications with limited computational resources.
• The approach demonstrates robustness across different imaging modalities, including medical images.