Robust dense registration of partial nonrigid shapes.

Area of Science:

• Computer Vision
• Geometric Processing
• Computational Geometry

Background:

• Dense registration of partial nonrigid shapes is crucial for various applications.
• Existing methods often struggle with accuracy, robustness, and efficiency.

Purpose of the Study:

• To present a complete and robust solution for dense registration of partial nonrigid shapes.
• To introduce a novel method based on heat kernel coordinates (HKCs) and priority-vicinity search.

Main Methods:

• Utilizing heat kernel coordinates (HKCs) for accurate point positioning and priority-based queuing.
• Employing a priority-vicinity search to ensure geometric compatibility during registration.
• Propagating dense registration progressively from detected feature correspondences.

Main Results:

• Achieved superior indexing ability, producing dense correspondences with fewer flips.
• Demonstrated robustness to noise and small holes due to the diffusion nature of HKCs.
• Significantly improved time performance through vicinity-based correspondence search.

Conclusions:

• The proposed method offers a technically sound and robust solution for dense registration of partial nonrigid shapes.
• HKCs and priority-vicinity search enable accurate, efficient, and flip-minimizing dense correspondences.
• The method's robustness makes it suitable for real-world applications without requiring surface repair.