Reconstructing Open Surfaces via Graph-Cuts.

Area of Science:

• Computer Graphics
• Computational Geometry
• Surface Reconstruction

Background:

• Reconstructing surfaces from unordered point sets is a fundamental problem in computer graphics and geometric modeling.
• Existing methods often struggle with open surfaces, nonorientable surfaces, and ensuring robustness and efficiency.

Purpose of the Study:

• To propose a novel graph-cuts-based method for robust and effective reconstruction of open surfaces from unordered point sets.
• To extend the method for handling more general surfaces, including nonorientable ones, using domain decomposition.
• To demonstrate the algorithm's parallel-friendly design and its ability to eliminate inter-domain issues.

Main Methods:

• A graph-cuts-based approach is employed, transforming the open surface problem into a watertight surface problem within a defined region using Boolean operations.
• Integration of variational models, Delaunay-based tetrahedral meshes, and multiphase techniques enhances reconstruction quality.
• A domain decomposition strategy is introduced for handling complex and nonorientable surfaces, with parallelization considerations.

Main Results:

• The proposed method successfully reconstructs open surfaces robustly and effectively from unordered point sets.
• The domain decomposition approach enables the handling of more general surfaces, including nonorientable types.
• Numerical examples validate the method's effectiveness on various surface types (watertight, open orientable, open nonorientable) and their combinations.

Conclusions:

• The novel graph-cuts-based method provides a robust and effective solution for open surface reconstruction.
• The domain decomposition extension broadens the applicability to complex and nonorientable surfaces.
• The algorithm's parallel-friendly design facilitates efficient computation for large datasets.