We have extended the work on stippling algorithms by introducing a scheme based on weighted centroidal Voronoi diagrams. The presented algorithm has very few user-specified parameters and requires no user interaction. In addition, the input data are greyscale images which can be produced by a wide variety of sources. Apart from simply requiring less work to generate a given stipple drawing, this independence allows cheap stippling to be used in a wider variety of situations than before.
The extension to precomputed stipple levels and thus real-time performance exacted a heavy cost in terms of visual quality, both in still images and in terms of inter-frame coherence of animations. The stipple levels are similar in concept to Praun et. al.'s Tonal Art Maps (TAMs) used to hatch 3D objects in . It should be straight-forward to generate high-quality stippling TAMs and use their approach to investigate frame-coherent animation. However, this would require abandoning our general image-based approach for 3D models.
The tone generated by a set of stipples is problematic and should be investigated further. All the results in this paper use a rational yet ad-hoc method to set the constant radii of the stipples in a particular image. A better understanding of the relationship between stipple radius, spacing and perhaps colour and the resulting perceived tone is required.
In addition, several interesting extensions to the current algorithm could be investigated, including varying the size of the stipples in a single drawing, and the use of color stipples. Partially transparent blended stipples could be used to aliveate some of the problems with stippling animations.
Adrian Secord 2001-11-20