Table of Contents
Fetching ...

Perceptually Uniform Construction of Illustrative Textures

Anna Sterzik, Monique Meuschke, Douglas W. Cunningham, Kai Lawonn

TL;DR

The paper quantifies how illustrative textures (stippling, hatching, triangles) are perceived, using crowdsourced pairwise comparisons and MDS to map their perceptual spaces. It finds that stippling and triangles align with a 1D manifold in a 2D space, while hatching splits into crosshatched and single-direction clusters and is only partially 1D. The authors reparameterize these spaces with Savitzky–Golay curve fitting and sigmoid density mappings to generate perceptually uniform texture levels, demonstrated through discrete and continuous data visualizations on surfaces. They illustrate practical applications in biomedical visualization and discuss limitations and avenues for future work, including exploring more texture types and 3D perception effects.

Abstract

Illustrative textures, such as stippling or hatching, were predominantly used as an alternative to conventional Phong rendering. Recently, the potential of encoding information on surfaces or maps using different densities has also been recognized. This has the significant advantage that additional color can be used as another visual channel and the illustrative textures can then be overlaid. Effectively, it is thus possible to display multiple information, such as two different scalar fields on surfaces simultaneously. In previous work, these textures were manually generated and the choice of density was unempirically determined. Here, we first want to determine and understand the perceptual space of illustrative textures. We chose a succession of simplices with increasing dimensions as primitives for our textures: Dots, lines, and triangles. Thus, we explore the texture types of stippling, hatching, and triangles. We create a range of textures by sampling the density space uniformly. Then, we conduct three perceptual studies in which the participants performed pairwise comparisons for each texture type. We use multidimensional scaling (MDS) to analyze the perceptual spaces per category. The perception of stippling and triangles seems relatively similar. Both are adequately described by a 1D manifold in 2D space. The perceptual space of hatching consists of two main clusters: Crosshatched textures, and textures with only one hatching direction. However, the perception of hatching textures with only one hatching direction is similar to the perception of stippling and triangles. Based on our findings, we construct perceptually uniform illustrative textures. Afterwards, we provide concrete application examples for the constructed textures.

Perceptually Uniform Construction of Illustrative Textures

TL;DR

The paper quantifies how illustrative textures (stippling, hatching, triangles) are perceived, using crowdsourced pairwise comparisons and MDS to map their perceptual spaces. It finds that stippling and triangles align with a 1D manifold in a 2D space, while hatching splits into crosshatched and single-direction clusters and is only partially 1D. The authors reparameterize these spaces with Savitzky–Golay curve fitting and sigmoid density mappings to generate perceptually uniform texture levels, demonstrated through discrete and continuous data visualizations on surfaces. They illustrate practical applications in biomedical visualization and discuss limitations and avenues for future work, including exploring more texture types and 3D perception effects.

Abstract

Illustrative textures, such as stippling or hatching, were predominantly used as an alternative to conventional Phong rendering. Recently, the potential of encoding information on surfaces or maps using different densities has also been recognized. This has the significant advantage that additional color can be used as another visual channel and the illustrative textures can then be overlaid. Effectively, it is thus possible to display multiple information, such as two different scalar fields on surfaces simultaneously. In previous work, these textures were manually generated and the choice of density was unempirically determined. Here, we first want to determine and understand the perceptual space of illustrative textures. We chose a succession of simplices with increasing dimensions as primitives for our textures: Dots, lines, and triangles. Thus, we explore the texture types of stippling, hatching, and triangles. We create a range of textures by sampling the density space uniformly. Then, we conduct three perceptual studies in which the participants performed pairwise comparisons for each texture type. We use multidimensional scaling (MDS) to analyze the perceptual spaces per category. The perception of stippling and triangles seems relatively similar. Both are adequately described by a 1D manifold in 2D space. The perceptual space of hatching consists of two main clusters: Crosshatched textures, and textures with only one hatching direction. However, the perception of hatching textures with only one hatching direction is similar to the perception of stippling and triangles. Based on our findings, we construct perceptually uniform illustrative textures. Afterwards, we provide concrete application examples for the constructed textures.
Paper Structure (13 sections, 2 equations, 14 figures, 2 tables)

This paper contains 13 sections, 2 equations, 14 figures, 2 tables.

Figures (14)

  • Figure 1: Subset of the stippling (first row) and triangle (second row) stimuli used in the perceptual studies. The images have been cropped and scaled down. The whole stimuli range, ranges from empty (0) to completely covered (1) and is sampled linearly with a step size of 0.05.
  • Figure 2: Hatching stimuli used in the perceptual studies. The images have been cropped to fit the page. In the experiment, an additional black texture (full density, completely covered) was present.
  • Figure 3: Example task of the hatching study. The participants rated the difference from 1 (very similar) to 9 (very different).
  • Figure 4: Scree plot for stippling, hatching, and triangles. Two dimensions seem to be sufficient to describe all three kinds of data.
  • Figure 5: Perceptual Spaces.
  • ...and 9 more figures