Fast Texture Mapping Adjustment via Local/Global Optimization.

Area of Science:

• Computer Graphics
• Geometric Modeling
• Computational Imaging

Background:

• Traditional texture mapping methods often struggle with visible seams and misalignment, especially at depth or texture boundaries.
• Model parameterization-based projective texture mapping can be computationally intensive and prone to distortions.

Purpose of the Study:

• To develop an efficient image-space texture optimization scheme for triangular mesh models.
• To reduce visible seams and misalignment in texture mapping.
• To accelerate the texture mapping process without compromising quality.

Main Methods:

• An image-space texture optimization scheme is proposed, differing from traditional model parameterization approaches.
• The method employs a local, parallel texture adjustment at boundaries, followed by a global correction step.
• The phased optimization addresses distortions caused by local adjustments.

Main Results:

• The novel scheme achieves a significant speed-up: 50-100x on GPU and 6x on CPU compared to state-of-the-art methods.
• Texture quality is maintained despite the substantial increase in processing speed.
• The approach demonstrates improved resilience to modeling and calibration errors.

Conclusions:

• The developed texture optimization scheme offers a fast and high-quality solution for texture mapping triangular mesh models.
• The method facilitates the automatic creation of textured 3D models using commodity depth sensors, even for untrained users.
• This work advances texture mapping techniques by improving efficiency and robustness.