CHS: Medium: Collaborative Research: Fast Photorealistic Computer Graphics Rendering of Non-Smooth Surfaces

CHS：媒介：协作研究：非光滑表面的快速真实感计算机图形渲染

• 批准号: 1703957
• 负责人: Ravi Ramamoorthi
• 金额: $ 40万
• 依托单位: University of California-San Diego
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-09-01 至 2022-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1703957&HistoricalAwards=false
• 关键词: CHS; Medium; Collaborative; Research; Fast

Computer graphics is finding growing application to virtual visual prototyping for the design of materials and products where appearance is important. Using graphics rendering on a display to replace physical prototypes could potentially make it faster and cheaper to bring new products and designs to market. But such applications demand high fidelity in the appearance of specific real materials on displays with ever-higher resolution. The fundamental problem is that current mathematical models for light reflection assume that details smaller than a pixel can be ignored, and replaced by smooth averages. At high resolutions this assumption breaks down; most materials start to appear noisy, grainy, or glittery as one looks closer. But to simply include microscopic detail in the model and continue with conventional rendering methods would be far too slow. This project will develop a range of methods to model the visible effects of sub-pixel details without greatly increasing the time and expense of rendering images. The results of this research will generalize a broad range of material models used in computer graphics so that they work under close-range observation. It will transform the field by fundamentally changing the definition of surface reflectance and by providing a suite of new tools for implementing and designing reflectance in industrial applications including automotive design, virtual prototyping, visual effects, and predictive product visualization.Accurately rendering the appearance of materials has always been a central problem of computer graphics. With today's ever-higher display resolutions, and applications demanding exact reproduction of specific materials, the status quo in material modeling is reaching its limits. The problem is that the standard approach to modeling surface reflectance, the Bi-directional Reflectance Distribution Function (BRDF), fundamentally assumes that surface roughness is far smaller than the scale of pixels. But rough surfaces, which can be modeled by smooth BRDF models for more distant or lower-resolution views, start to appear noisy, grainy, or glittery as one looks closer and the effects of individual scattering elements become visible. This produces "shimmering" or "glints" in a variety of materials that are difficult to model using current technology, including metallic paints used on cars, bead-blasted and brushed metal finishes popular for electronics, and the ubiquitous textured finishes on injection-molded plastic, as well as in fabrics, wood finishes and wood grain, and many other materials. This project will develop new ways to think about surface reflection in terms of reflectance models that build in spatial and angular variation at their core, without assuming smoothness at small scales. These models account for the visible effects of individual scattering features, and spatial and angular detail emerges naturally as the viewing and illumination conditions reveal it. Several mathematical representations for detailed materials are proposed, covering both surface and subsurface effects, to form the core of a full end-to-end pipeline that spans new acquisition methods and rendering techniques.

计算机图形学正在发现越来越多的应用虚拟视觉原型设计的材料和产品的外观是很重要的。在显示器上使用图形渲染来取代实物原型，可能会使新产品和设计推向市场的速度更快、成本更低。但这类应用要求在分辨率越来越高的显示器上显示特定真实材料的高保真度。最根本的问题是，当前的光反射数学模型假设小于一个像素的细节可以忽略，用平滑的平均值代替。在高分辨率下，这种假设就不成立了；大多数材料开始出现噪音，颗粒状，或闪烁，当一个人看得更近。但是，简单地在模型中包含微观细节并继续使用传统的渲染方法将会太慢。该项目将开发一系列方法来模拟亚像素细节的可见效果，而不会大大增加渲染图像的时间和费用。这项研究的结果将推广计算机图形学中广泛使用的材料模型，以便它们在近距离观察下工作。它将从根本上改变表面反射率的定义，并提供一套新的工具来实现和设计工业应用中的反射率，包括汽车设计、虚拟原型、视觉效果和预测性产品可视化。准确地呈现材料的外观一直是计算机图形学的核心问题。随着当今越来越高的显示分辨率，以及要求精确再现特定材料的应用，材料建模的现状正在达到极限。问题在于，模拟表面反射率的标准方法，即双向反射分布函数（BRDF），从根本上假设表面粗糙度远小于像素的尺度。但是粗糙的表面，可以用光滑的BRDF模型进行建模，用于更远或更低分辨率的视图，当人们看得更近时，就会开始出现嘈杂、颗粒状或闪闪发光的现象，并且单个散射元素的影响变得可见。这会在各种材料中产生“闪烁”或“闪光”，这些材料使用当前技术很难建模，包括汽车上使用的金属漆，电子产品中流行的珠喷和拉丝金属饰面，以及注塑塑料上无处不在的纹理饰面，以及织物，木材饰面和木纹，以及许多其他材料。该项目将开发新的方法，根据反射率模型来思考表面反射，这些模型在其核心处建立空间和角度变化，而不假设小尺度上的平滑。这些模型解释了单个散射特征的可见影响，空间和角度细节随着观察和照明条件的揭示而自然出现。提出了几种详细材料的数学表示，涵盖了表面和地下效果，形成了一个完整的端到端管道的核心，涵盖了新的获取方法和渲染技术。

项目成果

Rendering Neural Materials on Curved Surfaces

• DOI: 10.1145/3528233.3530721
• 发表时间: 2022-07
• 期刊: ACM SIGGRAPH 2022 Conference Proceedings
• 作者: Alexandr Kuznetsov;Xuezheng Wang;Krishna Mullia;Fujun Luan;Zexiang Xu;Miloš Hašan;R. Ramamoorthi