Virtual Material Acquisition and Representation for Computer Graphics. Dar'ya Guarnera
Чтение книги онлайн.

Читать онлайн книгу Virtual Material Acquisition and Representation for Computer Graphics - Dar'ya Guarnera страница 3

СКАЧАТЬ Preface

      The quest for photorealism drives research toward new developments in Computer Graphics and related fields, with the aim of finally conquering the uncanny valley and making the virtual world indistinguishable from the real one. Achieving a photorealistic virtual representation of the real-world involves several factors, among which is a faithful reproduction of material appearance. Human perception of the appearance of an object depends on the way its surface reflects the incident light, since the reflection provides important visual cues about material properties, such as glossiness and color, used to identify the type of material (plastic, metal, rubber, ceramic, fabric, etc.) and additional information, for instance dry/wet or smooth/rough textures. Visual properties are related to physical characteristics like the diffuse and specular albedo, the index of refraction of the material (as well as that of the physical media in which it is placed, such as air or water), its surface roughness, and so on. The interaction between an opaque, homogeneous surface and the lighting incident on it can be described by the radiometric Bidirectional Reflection Distribution Function (BRDF). A huge body of computer graphics research has been carried out in order to derive compact and accurate models for BRDFs.

      Selecting a suitable reflectance model to render a virtual material is not a straightforward task, since each reflectance model often aims to represent specific (subsets of) properties, with the result that a given model can describe plastic very well but not metals, and so on. Not all the BRDF models have become widespread in software packages, and the most common ones can be implemented differently in different renderers, leading to a slightly different rendered image. Hence, in this book we will focus on the theory rather than in the implementation. The reader will find accurate information about BRDF models, their taxonomy and characteristics, and the setup used to acquire the BRDF data, which range from low-to-high cost and low-to-high accuracy.

      Dar’ya Guarnera and Giuseppe Claudio Guarnera

      December 2017

       Acknowledgments

      This book is the product of several years of research on Virtual Material Appearance in Computer Graphics, ranging from BRDF acquisition to representation.

      We would like to acknowledge people who contributed to our research. First, we would like to express our gratitude to Dr. Mashhuda Glencross, who encouraged and supported us. We are grateful to Dr. Abhijeet Ghosh for the valuable suggestions on current BRDF models and acquisition setups.

      A heartfelt thank you to our parents, Alfio, Maria, Alexandr, and Nelli, and our sister Rosalinda and brother Sergey for their love, support, and encouragement. Last but not least a special thanks goes to Anastasiya and Ashanti for all the “input” in the development of this book.

      Dar’ya Guarnera and Giuseppe Claudio Guarnera

      December 2017

      CHAPTER 1

       Introduction

      The visual appearance of an object is a complex phenomenon, and to describe it properly it is important to understand how a material interacts with the light. For this purpose, many reflectance functions have been investigated, not only in computer graphics, where it is an active research field aiming to obtain photorealistic renderings, but many other disciplines like physics, optical engineering, computer science, and psychology, where considerable time and resources are being invested in the acquisition and representation of reflectance functions. In fact, material appearance plays an important role in many areas of science and industry:

      • Computer Vision (e.g., in object recognition applications)

      • Aerospace (e.g., for optimal definition of satellite mirrors reflectance and scattering properties)

      • Optical Engineering, Remote-Sensing (e.g., land cover classification, correction of view and illumination angle effects, cloud detection, and atmospheric correction)

      • Medical Applications (e.g., diagnostics)

      • Art (e.g., 3D printing)

      • Applied Spectroscopy (e.g., physical condition of a surface)

      • Film, Games, Virtual Reality, Marketing, etc.

      The world of technology is evolving and offering new, innovative technology that allows the “synthetic world” to look more and more realistic: as a consequence, digital reproduction of real-world material is growing rapidly. These results are driven by the combined effort of researchers, engineers and artists. However, despite the number of material reproduction techniques, material modeling is still a popular topic in research and industry since there is no straightforward way to digitize materials, and often the acquired data is not consistently reusable.

      As mentioned, a challenge in computer graphics is how to handle visual appearance accurately measuring and representing material characteristics from the real-world in order to replicate a material behavior and its interaction with the light. Several models and acquisition techniques suggested by researchers in recent years are often aimed at a particular subset of material properties. Ultimately, this limits the applicability of a method only to a specific class of materials. In fact, many methods and setups can properly deal with only a few classes of materials (e.g., leather, fabric, car paint, wood, plastic, rubber, mirrored surfaces, etc.) and unfortunately there is no universal way to acquire and represent all classes. If we look at the material representation side, it could be challenging to choose which a model. On the other hand, digital artists have been provided with applications that include material models with intuitive control, but generally they offer only few material models clearly identifiable with a known analytic formula in the scientific literature. Wider choice of material models is also available in physically based renderers, but those might be difficult to use and far from intuitive. An additional challenge is that the same material rendered in different applications might appear differently due to light tracing algorithms and other implementation choices. The cost of the material digitization could be high or low, where a high cost is not necessarily better. On the material acquisition side, time and costs often represent additional constraints, since they vary enormously across the available measurement devices, and high acquisition time and cost do not necessarily mean “better” (or worse, for that matter).

      A major challenge in computer graphics is how to simply and accurately measure the appearance of material characteristics from real-world objects and implement practical editable synthetic materials accurately matching the appearance of the original. Currently, no up-to-date universal material model that can represent leather, fabric, car paint, wood, plastic, rubber, mirrored surfaces, etc. exists [SDSG13]. A variety of rendering algorithms are used in the software pipeline, resulting in a need for optimized material representations, which requires both a flexible acquisition process and representation methods. Unfortunately, the following challenges still persist:

      • there is no widely adopted solution;

      • few solutions acquire material models that are good enough for a wide range of commercial applications without significant lor and money;

      • there is no standardized material model formats from acquisition setups;

      • there is little standardization across renderers, with different renderers supporting subsets of material properties;

      • material models are hard to edit by artists;

      • СКАЧАТЬ