FROST-BRDF: A Fast and Robust Optimal Sampling Technique for BRDF Acquisition

Area of Science:

• Computer Graphics
• Material Science
• Signal Processing

Background:

• Bidirectional Reflectance Distribution Function (BRDF) acquisition is computationally intensive, requiring millions of light-view samples.
• Existing methods face challenges in balancing acquisition speed with reconstruction accuracy.
• Compressed sensing offers a potential framework for reducing sampling requirements.

Purpose of the Study:

• To develop a novel technique for accelerating BRDF acquisition.
• To formulate BRDF acquisition as an optimal sub-sampling problem within compressed sensing.
• To minimize reconstruction error by designing a provably optimal sampling operator.

Main Methods:

• Formulated BRDF acquisition as a compressed sensing problem.
• Proposed the Fast and Robust Optimal Sampling Technique (FROST) for designing an optimal sub-sampling operator.
• Utilized a sparse representation formulation under the Multiple Measurement Vector (MMV) signal model.
• Converted an intractable combinatorial problem into a solvable optimization problem.

Main Results:

• FROST provides strong theoretical guarantees from compressed sensing.
• Demonstrated significant advantages in reconstruction quality over state-of-the-art methods via 10-fold cross-validation on public BRDF datasets.
• Achieved at least two orders of magnitude speedup compared to prior art.

Conclusions:

• FROST offers a computationally efficient and accurate method for BRDF acquisition.
• The technique is conceptually simple, easy to implement, and yields consistent results.
• This approach significantly advances the field of material appearance capture.