G&V: Medium: Collaborative Research: A Unified Approach to Material Appearance Modeling

G

• 批准号: 1064410
• 负责人: James Ferwerda
• 金额: $ 39.88万
• 依托单位: Rochester Institute of Tech
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2011
• 资助国家: 美国
• 起止时间: 2011-06-01 至 2016-05-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1064410&HistoricalAwards=false
• 关键词: G& Medium; Collaborative; Research; Unified

Realistic image synthesis techniques from computer graphics enable the use of simulation in a wide variety of important fields including architecture, industrial design and communication, military, medical, and emergency training, cultural heritage preservation, film production, and gaming. Realistic modeling of material appearance is an essential component of the image synthesis process. Current approaches to material modeling include analytical modeling, numerical simulation, and image-based capture. Each approach has distinct advantages and limitations, and different ranges of applicability. The lack of unity makes material modeling difficult and has limited the useful application of computer graphics image synthesis. This transformative research will change the way materials are modeled in computer graphics systems. Rather than using disparate models as at present, this project will unify these approaches into a common physical and perceptual framework that will serve as the basis for a rich set of tools for material modeling that are physically accurate, phenomenologically expressive, computationally efficient, and easy to use. This work should enable the use of computer-aided material design methods in a wide range of economically and culturally important applications. Creating this framework will involve three subprojects.Development of a material simulation testbed: In this subproject a suite of tools for material simulation will be developed that includes both Monte Carlo and deterministic algorithms. Different classes of materials (paints, metals, textiles) will be modeled, and different numerical methods will be tested and compared. The resulting simulation tools and a database of the simulated materials will be distributed.Unification of analytical, simulation, and image-based capture material modeling methods: In this subproject the analytical models that represent general classes of materials will be unified with simulation and image-based capture data that represent specific material instances. In the first part of this subproject simulation and capture data will be fit with a range of analytical models, considering both individual materials and "families" of materials generated by progressively changing the parameters of the simulation models. In the second part of this project methods for inferring the microstructures of materials measured using image-based capture methods will be developed. The approach will be to identify the class of a material and then vary the parameters of an appropriate simulation model to best reproduce the captured data. The results of this subproject will be expressive and efficient analytical material models that are physically grounded because they are based on captured data and rigorous simulations.Development of perceptually-based material design tools: An important criterion for material modeling is usability. Material designers need to be able to easily specify and visualize material appearance properties. This requires consideration of the human factors in material modeling. In this subproject a series of psychophysical experiments on material perception will be conducted and the results will be used to derive perceptually-based material models with meaningful parameters. How image properties affect the visual fidelity of rendered materials will also be investigated. These findings will then be used to develop effective and easy-to-use interfaces for computer-aided material design.Broader Impacts: Better methods for material modeling and rendering will lead to improved capability and productivity in fields such as architecture, industrial design and communication, training, cultural heritage, and entertainment. The project will build a material appearance community that stretches across academic and commercial boundaries to include computer graphics, computer vision and human vision researchers along with a range of industrial collaborators, and which focuses on developing effective solutions to real-world problems. The research will engage and train groups of students at 3 universities for scientific/technical careers that require working in interdisciplinary teams and partnering with coworkers in remote locations.

计算机图形学的逼真图像合成技术使模拟能够在各种重要领域中使用，包括建筑、工业设计和通信、军事、医疗和应急训练、文化遗产保护、电影制作和游戏。 材料外观的真实建模是图像合成过程的重要组成部分。 当前的材料建模方法包括分析建模、数值模拟和基于图像的捕获。 每种方法都有独特的优点和局限性，以及不同的适用范围。 缺乏统一性使得材料建模变得困难，并限制了计算机图形图像合成的有用应用。 这项变革性研究将改变计算机图形系统中材料建模的方式。 该项目不会像目前那样使用不同的模型，而是将这些方法统一到一个通用的物理和感知框架中，该框架将作为一组丰富的材料建模工具的基础，这些工具在物理上准确，在现象学上具有表现力，计算效率高且易于使用使用。 这项工作应该能够在广泛的经济和文化重要应用中使用计算机辅助材料设计方法。 创建该框架将涉及三个子项目。 材料模拟测试台的开发：在该子项目中，将开发一套材料模拟工具，其中包括蒙特卡罗和确定性算法。 将对不同类别的材料（油漆、金属、纺织品）进行建模，并测试和比较不同的数值方法。 由此产生的模拟工具和模拟材料的数据库将被分发。分析、模拟和基于图像的捕获材料建模方法的统一：在这个子项目中，代表一般材料类别的分析模型将与模拟和图像统一。基于代表特定材质实例的捕获数据。 在该子项目的第一部分中，模拟和捕获数据将与一系列分析模型相匹配，同时考虑到通过逐步改变模拟模型的参数而生成的单个材料和材料“家族”。 在该项目的第二部分中，将开发使用基于图像的捕获方法来推断材料微观结构的方法。 该方法将识别材料的类别，然后改变适当的模拟模型的参数，以最好地再现捕获的数据。 该子项目的结果将是富有表现力和高效的分析材料模型，这些模型具有物理基础，因为它们基于捕获的数据和严格的模拟。开发基于感知的材料设计工具：材料建模的一个重要标准是可用性。 材料设计师需要能够轻松指定和可视化材料外观属性。 这需要在材料建模中考虑人的因素。 在这个子项目中，将进行一系列关于物质感知的心理物理学实验，结果将用于导出具有有意义参数的基于感知的材料模型。 还将研究图像属性如何影响渲染材料的视觉保真度。 然后，这些发现将用于开发有效且易于使用的计算机辅助材料设计界面。更广泛的影响：更好的材料建模和渲染方法将提高建筑、工业设计和通信等领域的能力和生产力、培训、文化遗产和娱乐。 该项目将建立一个跨越学术和商业边界的材料外观社区，包括计算机图形学、计算机视觉和人类视觉研究人员以及一系列工业合作者，并专注于为现实世界问题开发有效的解决方案。 该研究将吸引和培训三所大学的学生群体从事科学/技术职业，这些职业需要在跨学科团队中工作并与偏远地区的同事合作。