CHS: Small: A wave optics foundation for predictive materials in computer graphics

点击翻译按钮获取中文摘要

Simulating how light reflects from surfaces is crucial to rendering realistic computer-generated images. Rendering is increasingly being used in applications, from product design to advertising to mass customization, that require representations of specific materials that are accurate enough to use for design and purchasing decisions, but current research on increasing accuracy and detail in reflection models is running up against fundamental limits of the ray optics approach that is traditionally used in computer graphics. This research aims to build a new foundation for materials in computer graphics based on the more accurate physics of wave optics, creating a new set of reflectance modeling tools that can be used to make existing models more accurate and to develop new models where they are needed. Renderings that can truly predict the appearance of new materials from their physical structure will have broad impact by enabling new and deeper integration of computer graphics into design, engineering, and manufacturing.New applications of computer graphics increasingly require models that can accurately predict light scattering based on information about the small-scale structure and optical properties of a material. In a great many cases these predictive models cannot be obtained using the traditional ray (geometric) optics approach, because materials have geometric structure at scales that overlap with the wavelengths of visible light. On the other hand, the traditional models developed for animation, visual effects, and games can effectively describe, though not predict, the appearance of many materials. These models need not be abandoned to achieve predictive appearance modeling, but principled ways to get from material descriptions to model parameters are needed. This work will put appearance models for graphics on a solid foundation of wave optics simulations that will clearly establish which materials can be accurately modeled using traditional models directly, which require adjustments to the parameters of those models, and which require new or generalized models to be able to render them adequately. Methods for finding model parameters to match given structures will be developed, as well as new models for phenomena that can't be described by existing ones.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.