open access

https://doi.org/10.14943/doctoral.k14626

https://hdl.handle.net/2115/82421

複雑な表面における反射の効率的なレンダリングに関する研究

英語

2021-06-30

Hokkaido University

ⅺ, 98p

Computer graphics has become a popular technique and is widely used in many applications including video games, virtual reality, and architectural designs. One of the important goals in these applications is to synthesize realistic images efficiently. Reflections on reflective surfaces with highly complexity structures, such as sparkles of the sun dancing on ocean surfaces or glints of street lights catching on wet pavements,are tiny but essential to the realism and quality of the synthetic images. The reflections visually describe the surfaces' detail and material to the viewer. However,these reflections are di cult to render accurately in a limited time. In Chapter 1, I explain the fundamentals and the relation between computer graphics and complex surfaces. Then, I discuss our objective and challenge in rendering the complex surfaces. Next, I present our approaches and their overall contributions. At the end of the chapter, I briefly explain my thesis outline and its structure, including each chapter's detail.This study is relying on many rendering and computing architectures. First, the rasterization-based rendering method excels at interactive rendering. Second,general-purpose computing on GPU is a high-performance parallel computing method. Last, the illumination model defines the calculation of interactions between light and surfaces. In Chapter 2, I explain this background knowledge in detail. The computer graphics research field has a long history of progressing in both result quality and process performance. There are also many research focusing on solving relevant topics to my work. In Chapter 3, I review the history of computer graphics research and the relevant research in multiple categories: reflection on smooth surfaces, glint and caustic, and anti-aliasing researches. I discuss the contributions and limitations of each study in detail in this chapter. Real-time glittery surface rendering is my first approach to solve complex surface reflection rendering in real-time. This work focuses on reproducing glint and sparkle on glittery reflective surfaces in an efficient manner. This work's main challenge is the fact that a reflection in each pixel is too complicated to be smoothly approximated. The proposed method efficiently renders the reflections from glittery surfaces on a static model and supports multiple light source types. The method stores numerous tiny surface orientations in one variable using slope space categorization and bitfield variables. In Chapter 4, I explain the method in detail, including its result and limitation. Real-time spatio-temporal reflection anti-aliasing is my second approach to the problem. This work aims at creating high-quality reflections of the Sun and Moon on moving water surfaces. The main challenges are complicated reflections, and its movement causes unstable reflections. The proposed method solves the problems by efficiently integrating reflections in-between the rendering frames to stabilize the result. The technique also approximates the water surfaces, that are too complicated to be rendered, by a smooth function for an efficient reflection computation. There are also two follow-up methods: adaptive supersampling and micro-reflection rendering. Adaptive supersampling improves my anti-aliasing method by preserving the sub-pixel geometric detail removed by the smooth function. Micro-reflection rendering increases the reflection quality of a simple surface, which the above process did not consider yet. In Chapter 5, I explain the primary method and the follow-up methods in detail. Then, I show its result and discuss its contribution and limitation. My proposed methods successfully generate reflections on glittery surfaces and moving water surfaces in real-time. However, there are some limitations and undiscovered potential in our techniques. In Chapter 6, I summarize the contributions and limitations of the proposed methods then explain what I plan to improve in the future.

2021-06-30

甲第14626号

博士(情報科学)

(主査) 教授 土橋 宜典, 教授 荒木 健治, 教授 坂本 雄児, 教授 長谷山 美紀

情報科学研究科(メディアネットワーク専攻)

博士論文

Version of Record