The success of graphics systems that render photorealistic output has obscured parallel developments in nonphotorealistic rendering. A review of these "expressive" systems suggests a framework for further development. One of the authors, Simon Schofield, has developed a prototype system that gives users a wide variety of automatic rendering styles-from the almost photorealistic, through the styles resembling artists' sketches or paintings, to abstract renderings of entirely new forms. The significant problems associated with NPR are primarily aesthetic and only secondarily technical. The culturally oriented theoretical problems surrounding the field are many and, at their most difficult, broach central issues of perception and representation. We therefore begin by discussing the limitations of photorealism as a representational style. However, the technical problems of implementing an NPR system are not trivial. As far as we know, the literature defines no general framework for an NPR system, although it does include an array of more focused solutions. We survey more than 15 of these solutions. On the basis of this survey, we then define a framework for NPR systems that distinguishes them from systems focused on photorealism. We also identify the characteristics of NPR systems that go beyond the tendency to simulate traditional media, functioning instead as 3D renderers. In this context, we briefly describe Schofield's system, called Piranesi, which operates this way.
@Article{ lansdown:1995:RNPR,
author = "John Lansdown and Simon Schofield",
title = "Expressive Rendering: A Review of Nonphotorealistic
Techniques",
journal = "IEEE Computer Graphics and Applications",
volume = "15",
number = "3",
pages = "29-37",
month = "May",
year = "1995",
}