A framework for realistic image synthesis 合成. Proceedings of SIGGRAPH 1997

https://www.cis.rit.edu/people/faculty/ferwerda/publications/framewrk.pdf
abstract
our goal is to develop physically based lighting models and perceptually based rendering procedures for computer graphics that will produce synthetic images that are visually and measurably indistinguishable from real-world images. Fidelity of the physical simulation is of primary concern. our research framework is subdivided into three sub-sections: the local light reflection model, the energy transport simulation, and the visual display algorithms. the first two subsections are physically based, and the last is perceptually based.

we emphasize the comparisons between simulations and actual measurements, the difficulties encountered, and the need to utilize the vast amount of psychophysical research already conducted. future research directions are enumerated. we hope that results of this research will help establish a more fundmamental, scientific approach for future rendering algorithms. this presentation describes a chronology 万年历 of past research in global illumination and how parts of our new system are currently being developed.

1 introduction
from its infancy in the late 1960’s, the quality of computer graphics images have improved at fantastic rates. the initial renderings of simple environments with only direct lighting have been transformed into pictures of complex scenes with shadows, shading, and global interreflections. for several decades now, high quality simulations have been used for a large number of tasks such as pilot training, automotive design, and architectural walkthroughs[GREE91]. the entertainment industry has developed techniques for creating startling special effects and realistic simulations with dramatic results. even virtual reality games use convincing imagery with great success.

but are these images correct? would they accurately represent the scene if the environment actually existed? in general, the answer is no; yet the results are appealing because the resulting images are believable.

if we could generate simulations that were guaranteed to be correct, where the algorithms and resulting pictures were accurate representations, then the simulations could be used in a predictive manner. this could be a major paradigm shift for the computer graphics industry, but would have much broader applicability than just picture making.

a look at how accurate simulations are used in other areas might clarify this point. the entire electronics industry is now based on simulations for chip design芯片设计; these simulations are standardly used for testing and modifications prior to fabrication. in color science, we utilize the response matching functions for color transformations without recreating the matching experiments. why can not we use computer graphics algortihms for the testing and development of printing technologies, photographics image capture or display devices? why can not these accurate but artificial scenes be used for algorithmic development in image processing, robotics and machine vision? if we knew that the simulated images were correct, we could easily control and isolate the variables, obtain any projection or resolution desired, and avoid the difficulties and constraints of experimental measurements.

however, in order to be predictive, we must prove that the simulations are correct. fidelity is the key. this difficult task requires a major multi-disciplinary effort between physicists, computer scientists, and perception psychologists and also requires experimental measurements and comparisions. unfortunately to date there has been relatively little work done in correlating the results of computer graphics simulations with real scenes. however, with more accurate image acquisition and measurement devices, these comparisions will be achievable if we can generate adequate computer simulations.

from early computer generated images such as the phong goblet (1975) to the synthesized 合成的 pictures of today, there has been an exponential growth in the complexity of environments (figure 1). this increased complexity has also led to an exponential growth in computational costs for realistic rendering (figure 2). but the available processing power has also increased exponentially.
场景的复杂度在增加，计算机处理能力也在增加。

according to Moore’s Law, with a doubling of chip density every 18 months, we now have approximately 4000 times the processing power that was available when the first graphics algorithms were developed (figure 3). there has also been a concomitant 伴随的 increase in memoery capacity, which offsets the constraint on environment complexity, as well as a vast reduction in cost per compute cycle. as we look toward the future a combination of increasing computation power and algorithmic improvements will certainly allow us to compute images that are physically and perceptually correct.

the purpose of this monograph 专著 is to describe in general terms our long-term development efforts attempting to achieve these tasks, to describe the difficulties encountered, and to encourage the computer graphics community to develop physically based algorithms of great realism and fidelity. although there are many frontiers 前沿 for future research in computer graphics, for physically based realistic image synthesis, three areas are especially critical: local light reflection models, light transport simulation, and perceptually based issues.

our specific long term goals are:
light reflection models

1. the development of a general purpose wavelength-dependent model or models for arbitrary reflectance functions including the effects of subsurface scattering and re-emission, texturing and surface anisotropy.
2. validation of the local light reflection model through comparisons with measured physical experiments
3. a means for representing this model in a compact, data-efficient form useful for progressive rendering algorithms.
4. establishing and distribution of the reflectance characteristics of materials whose accuracy has been verfied by measurements.