Article Preview
TopIntroduction
Stippling is a technique for drawing, engraving, or painting using small dots or short strokes. Ideally, an automatic stippling method can convert an image input into a placement of dots with minimal user intervention; the resulting stippled image can depict the input image effectively. Computer-generated stippling methods can be broadly categorized as seeking tone-based, structure-based, or hand-drawn distributions, with most (Deussen et al., 2000; Hiller et al., 2003; Kopf et al., 2006; Schlechtweg et al., 2005; Secord, 2002; Vanderhaeghe et al., 2007) concentrating on preserving tone. Blue noise has generally been considered the ideal stipple distribution, with minimal variation in stipple spacing (Kopf et al., 2006; Secord, 2002). Structure-based stippling methods (Son et al., 2011; Kim et al., 2008; Kim et al., 2010; Mould, 2007; Martín et al., 2011) seek to preserve fine details in addition to tone matching. Both structure-aware and tone-based stippling methods, using the stipple distribution to express the image information, can seek either a halftoning effect or an illustrative effect, as opposed to simulating hand-drawn effects. Automatic hand-drawn stippling methods (Maciejewski et al., 2008; Kim et al., 2009; Martín et al., 2010) approximate the look of real stippling with irregular placement and shape variation. Other possible effects, such as heightening (DeCarlo & Rusinkiewicz, 2007), scratchboard (Yamamoto et al., 2004), or line drawing (Son et al., 2011), have received less attention.
Historically, stippling algorithms have been chiefly concerned with tone matching. As a consequence, many stippling results look similar to halftoning effects. However, stippling and halftoning are distinct, despite the common use of black stipples to approximate continuous tones. Deussen et al. (2000) characterize stippling in opposition to halftoning, saying that in stippling, “a smaller number of relatively large dots is used which vary in size and sometimes in shape”. We approach stippling from the halftoning perspective; our interest has been provoked by recent structure-focused halftoning approaches (Chang et al., 2009; Li & Mould, 2010; Pang et al., 2008), which provide excellent texture preservation as well as good tone matching, and we sought a stippling algorithm that has these properties. Our stipples have regular shape (always circular) and vary in size in a certain degree; for us, the spatial distribution of stipples is the critical issue. We not only match tone, but seek also to communicate structural details of the image with contrast awareness in the dot distribution. Figure 1 illustrates the point. With a very large stipple count (left), stipples can communicate image content by halftoning; with fewer stipples (right), tone is no longer preserved and stipples must be deployed to structural aspects in order to convey the image. We seek to provide a continuum between the examples in the figure.
Figure 1. Left: halftoning with stipples, by Flickr user cdslug. Right: illustrative stippling, by Flickr user makedonche19. Both images are within Creative Commons license