A Closed-Form, Pairwise Solution to Local Non-Rigid Structure-From-Motion.

Area of Science:

• Computer Vision
• 3D Reconstruction
• Geometric Modeling

Background:

• Non-Rigid Structure-from-Motion (NRSfM) commonly uses local, differential constraints between image pairs.
• Existing methods often involve computationally expensive polynomial solvers and can produce unreliable results due to degeneracies.

Purpose of the Study:

• To address the computational cost and reliability issues in isometric/conformal NRSfM.
• To develop a novel method for deriving and solving surface normal equations efficiently and accurately.

Main Methods:

• Derived a new system of equations for surface normals solvable in closed-form.
• Developed a local disambiguation mechanism for the two potential solutions.
• Introduced a reliability assessment to discard unreliable normal estimates.

Main Results:

• The new method provides two closed-form solutions for surface normals, easily disambiguated locally.
• The approach can reliably assess and discard inaccurate local normal estimates.
• Experimental results demonstrate significantly improved accuracy and speed compared to state-of-the-art NRSfM methods.

Conclusions:

• The proposed method offers a more computationally efficient and reliable solution for isometric/conformal NRSfM.
• It overcomes limitations of existing high-degree polynomial equation systems and local degeneracies.
• This advancement enables more robust and accurate 3D surface reconstruction from image sequences.