DCUDF2: Improving Efficiency and Accuracy in Extracting Zero Level Sets From Unsigned Distance Fields.

Area of Science:

• Computer Graphics
• Geometric Modeling
• Computational Geometry

Background:

• Unsigned distance fields (UDFs) are versatile for representing complex 3D models.
• Extracting accurate zero level sets from UDFs is difficult, often compromising topology and detail.
• Existing methods struggle with precision, robustness, and efficiency.

Purpose of the Study:

• To introduce DCUDF2, an improved method for extracting zero level sets from UDFs.
• To enhance geometric fidelity and preserve topological correctness in model extraction.
• To increase the robustness and runtime efficiency of UDF processing.

Main Methods:

• Developed an accuracy-aware loss function with self-adaptive weights for precise fitting.
• Implemented a topology correction strategy to reduce hyper-parameter sensitivity.
• Introduced new operations utilizing self-adaptive weights to speed up convergence.

Main Results:

• DCUDF2 demonstrates superior geometric fitting compared to prior methods.
• The method effectively preserves fine geometric details and topological correctness.
• Experiments show significant improvements in runtime efficiency and robustness.

Conclusions:

• DCUDF2 offers a robust and efficient solution for accurate zero level set extraction from UDFs.
• The enhanced approach overcomes limitations of previous methods in geometric and topological accuracy.
• This work advances the state-of-the-art in UDF-based 3D model representation and processing.