SGER: Perceptually-inspired Algorithms for Shape Processing and Abstraction

点击翻译按钮获取中文摘要

SGER: Perceptually-inspired Algorithms for Shape Processing and AbstractionDoug DeCarloRutgers UniversityAbstractWhen an artist draws a particular shape, they omit, distort and abstract various aspects of its structure to make it more recognizable and understandable. For instance, cartographers simplify shape details on maps to make them easier to use. In accomplishing this, artists use their own eyes to judge the effects of their decisions, in order to reach a successful depiction. Computers don't have this luxury and must approach this differently. This research develops a computational approach to processing and abstracting shape that is based upon models of human visual perception. Our system can thus make predictions about what shape the viewer will see, and can therefore make effective depictions of simplified or distorted shapes. Our approach processes shapes in terms of local symmetries, and organizes them into a collection of perceptual parts. Certain parts are removed and manipulated in order to produce the desired abstract depiction of the original shape.