How To Draw Velvet Texture

Figure 1.

Satin (left) is visually more similar to aluminum foil (middle) than to velvet (right). However, satin certainly belongs to a different material class than aluminum. The first two images were downloaded from Morguefile.com and the third image from pxfuel.com, released under free license.

Figure 2.

An example of each of the two conditions, within the interface. (Left) The full figure condition, in which the figure or object with the target fabrics is fully visible. On the right, the crop condition, where only a patch from the target fabric is visible, which is intended to deprive the visual system from context and shape information. Note that a participant would see only the left or right screen, never both. Gabriël Metsu, A Young Woman Composing a Piece of Music, 1664, Mauritshuis.

Figure 3.

Consistency within and between participants. (Left) The consistency within participants is calculated as the averaged pairwise correlation between each participants repetitions of the stimuli, and the error bars indicate the standard error. Right) The consistency between participants was calculated using intraclass correlations, and the error bars indicate the 95% confidence interval. The full report of the ICC analysis can be found in Table S1. Note that non-significant ICC are not visualized (i.e., hairiness in the crop condition).

Figure 4.

The perceptual judgments of satin and velvet, for both conditions. In the top plots, the data are split by viewing condition, whereas in the bottom plots data are divided by material. For each participant, we took the median rating across the stimuli repetitions, and then averaged across these values. Significance between condition (top) and material (bottom) is indicated at p < 0.05, Bonferroni corrected. Note that besides the significance, the top and bottom display the same data, only differently presented to make interpretations across conditions easier, and to avoid visual clutter of displaying all significant differences within a single plot.

Figure 5.

Correlation matrices of the attributes for both conditions. Color indicates the magnitude of the correlation coefficient. Asterisk (*) indicates a significant effect at p < 0.05.

Figure 6.

PCA biplot for the full figure condition. The materials are clustered within 95% confidence ellipses. Attributions of all stimuli can be found in Supplementary Figure S1 in the supplementary materials.

Figure 7.

PCA biplot for the crop condition. The materials are clustered within 95% confidence ellipses. Attributions of all stimuli can be found in Supplementary Figure S1 in the supplementary materials.

Figure 8.

Left: The mean ratings on the y-axis of the attributes for one specific stimulus. An asterisk (*) indicates a significant difference at p < 0.05. Right: the crop and full figure stimuli represented in the left bar chart. Error bars indicate the standard error. Anthony van Dyck, Catherine Howard, Lady d'Aubigny, 1638, National Gallery of Art.

Left: The mean ratings on the y-axis of the attributes for one specific stimulus. An asterisk (*) indicates a significant difference at p < 0.05. Right: the crop and full figure stimuli represented in the left bar chart. Error bars indicate the standard error. Anthony van Dyck, Catherine Howard, Lady d'Aubigny, 1638, National Gallery of Art.

Figure 9.

The original full figure stimuli, with red boxes that indicate the crops made for this stimulus. Each of the 19 stimuli from Experiment 1 was subdivided into a set of crops as shown here. These sets of crops were used as stimuli in experiment two. Each crop within a set was the same size. Anthony van Dyck, Portrait of Agostino Pallavicini, 1621, J. Paul Getty Museum.

The original full figure stimuli, with red boxes that indicate the crops made for this stimulus. Each of the 19 stimuli from Experiment 1 was subdivided into a set of crops as shown here. These sets of crops were used as stimuli in experiment two. Each crop within a set was the same size. Anthony van Dyck, Portrait of Agostino Pallavicini, 1621, J. Paul Getty Museum.

Figure 10.

(A) Luminance distribution and the highlight mode used as threshold value (black bar) to create the binarized image. (B) The original stimulus, as presented to the participants in the rating experiment, and its binarized version.

Figure 11.

The intrarater and interrater agreement. The top contains the intrarater agreement (consistency within observers) with shininess on the left and softness on the right. The same ordering is applied at the bottom for the interrater agreement (agreement between observers). The error bar indicates the standard deviation.

Figure 12.

Correlation coefficients of shininess (top) and softness (bottom) with the image features highlights' contrast, highlights' coverage, and mean luminance of the crops. The values reported are significant at p < 0.05.

Figure 13.

One crop that was identified as a possible outlier. With this stimulus included, a strong negative correlation was found between softness and roughness, which is surprising based on the literature. With this crop removed, the correlation is no longer significant. This might be due to the visibility of the individual brushstrokes, which gave rise to a perceptual ambiguity.

Figure 14.

Visualization of the crop set 2 from the bottom of Figure 12. The image on top shows the locations where the crops were taken from the whole fabric. The crop in the top row were perceived to be significantly softer than the crops in the bottom row. Anthony van Dyck, Catherine Howard, Lady d'Aubigny, 1638, National Gallery of Art.

Visualization of the crop set 2 from the bottom of Figure 12. The image on top shows the locations where the crops were taken from the whole fabric. The crop in the top row were perceived to be significantly softer than the crops in the bottom row. Anthony van Dyck, Catherine Howard, Lady d'Aubigny, 1638, National Gallery of Art.

Figure 15.

Examples from our stimuli set of a neat (left) and a loose use of brushstrokes (right), to illustrate the difference in these two styles which becomes apparent when moving closer to the physical painting—or when zooming in. The two crops (below) are from the crop-set that correspond to the paintings (above). Left: Adriaen van der Werff, Self-portrait with the Portrait of his Wife, Margaretha van Rees, and their Daughter Maria, 1699, Rijksmuseum. Right: Frans Hals, Portrait of a Man, Possibly Nicolaes Pietersz Duyst van Voorhout, ca. 1636–38, The Metropolitan Museum of Art.

Table 1.

The factor loadings for the first two principle components of two PCAs, one for each condition.

Table 2.

Results of One-Way ANOVAs of Experiment 2.