Indirect decoding edges for one-shot shape acquisition.

Area of Science:

• Computer Vision
• Computational Imaging
• 3D Reconstruction

Background:

• Accurate 3D shape reconstruction using structured light techniques is crucial.
• Decoding coded stripe patterns is a key challenge in one-shot 3D reconstruction.
• Existing methods often struggle with complex patterns and lack direct comparison benchmarks.

Purpose of the Study:

• To propose and evaluate an indirect decoding method for structured light patterns.
• To address the limitations of traditional edge-based decoding algorithms.
• To provide a quantitative comparison of different decoding methods using ground truth data.

Main Methods:

• Modeling stripe pattern decoding as matching two code sequences (edge, color, or mixed).
• Developing an indirect decoding approach that matches color or mixed code sequences.
• Generating six benchmark datasets using space-time analysis for ground truth.
• Quantitatively evaluating five decoding methods, including the proposed approach.

Main Results:

• The proposed indirect decoding method demonstrates superior performance compared to traditional methods.
• Indirect decoding offers advantages such as higher Hamming distance and enforced local coherence.
• The method proves robust in complex coding scenarios and outperforms state-of-the-art techniques.

Conclusions:

• Indirect decoding of structured light patterns is a more effective approach for 3D shape reconstruction.
• The developed benchmark datasets enable direct and quantitative comparison of decoding algorithms.
• This work advances the field of structured light 3D reconstruction, particularly for challenging scenarios.