Evaluating Non-Photorealistic Rendering

点击翻译按钮获取中文摘要

ABSTRACT0227337Douglas M. DeCarloRutgers U, New BrunswickWe propose to show that non-photorealistic renderings which exhibit meaningful abstractionare easily understood. Abstract images are designed to make meaningful structure clear. We haveworked on the interactive transformation of photographs into versions which manifest meaningfulabstraction by drawing on eye movement data [DeCarlo & Santella SIGGRAPH 2002, Santella &DeCarlo NPAR 2002]. Over the next year, we intend to carry out psychophysical experiments thatcould both validate our interactive technique, and provide new ways of learning more about humanvision eye movements, in particular.This is high-risk research. Non-photorealistic rendering (NPR) is an established field, but hasnever been evaluated. Psychophysical methods are typically applied to simple patterns (i.e. dotsand blobs). However, our method already involves a close link between NPR and perception andpsychophysics, and our lab maintains close connections with psychophysicists including EileenKowler and Thomas Papathomas (both are, like us, affiliated with the Rutgers Center for CognitiveScience). Many simple conjectures and experiments can be stated naturally in our framework; webelieve we can obtain compelling results.