Fabricating microgeometry for custom surface reflectance

Tim Weyrich, Pieter Peers, Wojciech Matusik, Szymon RusinkiewiczACM Trans. Graphics 2009 · DOI

TL;DR

This paper introduces a system to fabricate physical surfaces with user-defined reflectance properties by manufacturing microgeometry. By translating a desired highlight shape into a precise surface height field, the authors demonstrate the creation of custom materials using standard CNC milling machines, opening doors for novel material design in various fields.

What Problem Does This Paper Address?

In computer graphics, accurately modeling how light interacts with surfaces (their reflectance) is crucial for realistic rendering. While complex reflectance models (BRDFs) exist, the ability to physically realize surfaces with specific reflectance properties has been limited. This paper bridges the gap between digital material design and physical manufacturing by enabling the creation of custom surface microgeometry that dictates how light reflects, moving beyond simply selecting existing materials to actively designing them.

Methods

The authors propose a pipeline that starts with a user-defined target reflectance, often specified as a desired highlight shape. This shape is then used to infer the necessary distribution of microfacet slopes on the surface. The core of their method involves:

  • Microfacet Distribution Inference: They convert the user's highlight shape into a probability distribution of microfacet orientations. This involves a deconvolution step using the Lucy-Richardson algorithm to account for the base material's inherent reflectance.
  • Sampling and Optimization: The microfacet distribution is sampled using a low-discrepancy technique (similar to image stippling) to ensure good coverage. These sampled microfacets are then arranged into a tileable height field using a simulated-annealing optimization process. This optimization aims to minimize discontinuities and valleys, ensuring manufacturability.
  • Height Field Generation: A discrete Poisson equation is solved to determine the optimal height for each microfacet, resulting in a smooth, continuous surface.
  • Manufacturing: The generated height field is then translated into GCode for a computer-controlled milling machine to physically fabricate the surface.

Datasets

Not discussed in the paper. The authors generate their own height fields based on user input and measured BRDFs.

Key Insights and Ideas

  • Highlight Shape as Design Input: The idea that a simple 2D highlight shape can serve as a powerful and intuitive input for designing complex 3D surface reflectance is a key insight.
  • Bridging Digital and Physical: The paper demonstrates a practical workflow for translating digital surface designs into tangible, physically manufactured materials, which has significant implications for design and fabrication.
  • Optimization for Manufacturability: The emphasis on optimizing for continuity and minimizing valleys is crucial for making the generated designs feasible with real-world milling machines, acknowledging the constraints of fabrication.
  • Microgeometry is Key: The work reinforces the understanding that subtle variations in surface geometry at the micro-level are the primary drivers of macroscopic surface appearance.

Key Results

The paper successfully demonstrates the fabrication of several custom surfaces with user-defined reflectance properties. This includes:

  • Reproductions of measured BRDFs (e.g., brushed aluminum).
  • Materials with unconventional highlights, such as spiral patterns and starbursts.
  • The resulting milled surfaces exhibit the desired reflectance characteristics when illuminated, as verified by simulations and photographs of reflections.
  • The authors show that their optimization process, particularly the inclusion of a valley penalty, significantly reduces unwanted artifacts like central highlights compared to not using it.

Limitations and Future Work

The authors acknowledge several limitations:

  • Resolution: The current system uses a 30x30 grid of microfacets, which leads to poor sampling of large uniform areas, visible in some of the results (e.g., the bat logo).
  • Base BRDF Limitations: The system assumes a spatially homogeneous and opaque base BRDF, limiting the range of achievable reflectance properties.
  • Tool Shape Artifacts: The non-zero extent of the milling bit creates spurious horizontal strips in valleys and can lead to unwanted central highlights, even with optimization.
  • Spurious Highlights: Cross-shaped highlights are observed, attributed to remaining vertical discontinuities and the inability to mill perfectly vertical surfaces.

Future work suggested includes:

  • Manufacturing larger patches of microfacetted surfaces and exploring tiling for observable highlights on the surface itself.
  • Investigating curved surfaces to enhance highlight variation.
  • Exploring alternative manufacturing technologies (laser engraving, chemical etching, water-jet milling) that might offer higher precision and a wider range of base materials.
  • Developing methods to handle anisotropic base materials.
  • Accounting for shadowing, masking, and interreflection in more complex scenarios for a fuller inverse-rendering problem.

Summary updated

Abstract

We propose a system for manufacturing physical surfaces that, in aggregate, exhibit a desired surface appearance. Our system begins with a user specification of a BRDF, or simply a highlight shape, and infers the required distribution of surface slopes. We sample this distribution, optimize for a maximally-continuous and valley-minimizing height field, and finally mill the surface using a computer-controlled machine tool. We demonstrate a variety of surfaces, ranging from reproductions of measured BRDFs to materials with unconventional highlights.
BibTeX
@article{weyrich2009fabricating,
  author = {Tim Weyrich and Pieter Peers and Wojciech Matusik and Szymon Rusinkiewicz},
  title = {{Fabricating microgeometry for custom surface reflectance}},
  journal = {ACM Transactions on Graphics},
  year = {2009},
  doi = {10.1145/1531326.1531338},
}
Figure