Why do rough surfaces appear glossy?

Lin Qi, Mike J. Chantler, J. Paul Siebert, Junyu DongJournal of the Optical Society of America A 2014 · DOI

TL;DR

This paper challenges the common assumption that rougher surfaces are always glossier. Through controlled experiments, the authors demonstrate that while some roughness can increase perceived gloss, excessive roughness actually reduces it, leading to a non-monotonic relationship. They identify the "percentage of highlight area" as a key image statistic strongly correlated with perceived gloss, suggesting it's a crucial cue for how we perceive surface shininess.

What Problem Does This Paper Address?

For decades, research on gloss perception has largely focused on smooth surfaces like spheres. While some studies have suggested that increasing the roughness of more complex surfaces leads to increased perceived gloss, this paper questions this simple, monotonic relationship. The authors aim to understand how mesoscale roughness affects gloss perception, particularly under more realistic rendering and illumination conditions, and to identify the underlying visual cues responsible for these perceptions. They hypothesize that the "illusory gloss" observed on rough surfaces under specific, simplified lighting might not hold true in more naturalistic scenarios.

Methods

The researchers employed a two-experiment approach using computer-generated 1/f<sup>β</sup> noise surfaces to model mesoscale roughness.

  • Surface Modeling: 1/f<sup>β</sup> noise was used to create height maps, where the parameter β controls the roughness. Higher β values correspond to smoother surfaces, while lower β values indicate rougher surfaces.
  • Rendering: A physics-based path tracer ("LuxRender") was used for rendering. Crucially, they compared simple single-bounce rendering with more realistic multiple-bounce path tracing. They also utilized a high dynamic range (HDR) environment map ("StPeters") for illumination, simulating more natural lighting conditions, as opposed to simpler point or area light sources.
  • Experiment 1: Investigated perceived gloss of Lambertian surfaces with varying RMS height (σ) under different rendering conditions (frontal, oblique, and environment lighting, with single and multiple bounces).
  • Experiment 2: Explored the effect of varying mesoscale roughness (controlled by β) on perceived gloss using a gloss BRDF model and realistic HDR illumination. To provide motion parallax cues, the surfaces were animated by rotating them.
  • Image Statistics Analysis: After collecting subjective gloss ratings, the researchers analyzed several image statistics related to specular highlights, including percentage area, average size, strength, number, and spread. They developed a simple model to explain the relationship between roughness and the "percentage of highlight area."

Datasets

The paper does not introduce new datasets. Instead, it utilizes:

  • Synthetic Datasets: 1/f<sup>β</sup> noise surfaces generated with varying parameters (RMS height σ and roll-off factor β).
  • Illumination Map: The "StPeters" HDR environment map from Debevec's Light Probe Image Gallery was used to simulate realistic lighting.

Key Insights and Ideas

  • Realistic Rendering is Crucial: The perceived gloss of rough surfaces is highly dependent on the rendering conditions. Simple frontal illumination can create an "illusory gloss" on rough Lambertian surfaces, which disappears under more naturalistic HDR environment lighting.
  • Non-Monotonic Gloss-Roughness Relationship: Contrary to some previous findings, increasing roughness beyond a certain point reduces perceived gloss. This suggests a unimodal, non-monotonic relationship between mesoscale roughness and perceived gloss.
  • "Percentage of Highlight Area" as a Key Cue: The percentage of pixels identified as specular highlights (using a global intensity threshold) emerged as the strongest predictor of perceived gloss, accounting for 81% of the variance in their data. This suggests that the extent of specular reflection, rather than just its presence, is a critical factor.
  • Geometric Explanation for Reduced Gloss: The authors propose a model where the distribution of surface slope angles interacts with the distribution of bright areas in the illumination environment. For very rough surfaces, a larger proportion of facets may be oriented such that they are self-occluded or cannot "see" the bright light sources, thus reducing highlight formation and perceived gloss.

Key Results

  • Experiment 1: Perceived gloss increased with RMS height under frontal illumination but decreased with oblique lighting. Under the HDR environment map, perceived gloss was significantly lower, and inter-reflections had no significant effect.
  • Experiment 2: Perceived gloss initially increased with increasing roughness (decreasing β) but then rapidly decreased after reaching a peak around β ≈ 1.9. This confirmed the non-monotonic relationship.
  • Image Statistics: The "percentage of highlight area" (Pha) showed a strong positive correlation (ρ = 0.90, p < 0.001) with perceived gloss. A simple model predicting Pha based on surface slope angle distribution and illumination distribution also showed a significant correlation (ρ = 0.64, p < 0.05) with the observed Pha.

Limitations and Future Work

  • Limitations:
    • The definition of specular highlights used in their image statistics analysis was a simplistic global intensity threshold, and they acknowledge that more sophisticated methods for highlight segmentation exist.
  • The BRDF model used (Ashikhmin-Shirley) might not be ideal for wide-angle gloss, and they suggest exploring a wider range of BRDF models in future work.
  • The explanation for the "illusory gloss" in frontal illumination (Experiment 1) is not fully detailed, suggesting it might be related to second-order effects.
  • Future Work:
    • Investigate the effect of varying illumination distributions on perceived gloss of complex surfaces.
  • Explore a wider range of BRDF models to better capture different types of gloss.
  • Further refine the model explaining the relationship between surface slope angles, illumination distributions, and highlight formation.

Summary updated

Abstract

The majority of work on the perception of gloss has been performed using smooth surfaces (e.g., spheres). Previous studies that have employed more complex surfaces reported that increasing mesoscale roughness increases perceived gloss [Psychol. Sci.19, 196 (2008), J. Vis.10(9), 13 (2010), Curr. Biol.22, 1909 (2012)]. We show that the use of realistic rendering conditions is important and that, in contrast to [Psychol. Sci.19, 196 (2008), J. Vis.10(9), 13 (2010)], after a certain point increasing roughness further actually reduces glossiness. We investigate five image statistics of estimated highlights and show that for our stimuli, one in particular, which we term "percentage of highlight area," is highly correlated with perceived gloss. We investigate a simple model that explains the unimodal, nonmonotonic relationship between mesoscale roughness and percentage highlight area.
BibTeX
@article{qi2014why,
  author = {Lin Qi and Mike J. Chantler and J. Paul Siebert and Junyu Dong},
  title = {{Why do rough surfaces appear glossy?}},
  journal = {Journal of the Optical Society of America A},
  year = {2014},
  doi = {10.1364/josaa.31.000935},
}
Figure