Vol.33 No.3 August 1999

A Look at the 3D Graphics Industry

Scott S.Fisher
Telepresence Research Inc.
Glen Fraser

August 99 Columns
Images and Reversals CG Around the World

Real-Time Interactive Graphics
Previous Real-Time Interactive Graphics Next Real-Time Interactive Graphics

In this month’s Real-time Interactive Graphics column we take a look at the 3D graphics industry through the unique perspective of long-time industry analyst and observer, John Latta. Latta’s company, 4th Wave, Inc. publishes a widely read newsletter on 3D and digital media called the Wave Report.

A Reminder

We are very interested in hearing from our readers. In particular, if you are involved in a new (real-time, interactive) application, or have a related topic that you think might be interesting to share with the SIGGRAPH membership through our column, please let us know. Also, if you have feedback to offer on current or past columns, we would be more than glad to hear it!

— Glen D. Fraser, Scott S. Fisher

3D: To Z or Not to Z

John Latta
4th Wave, Inc.

3D technology is at an apex of adoption and visibility. The disappearance of the role of 3D technology from Star Wars Episode I is taken for granted in much the same way that it is assumed that a new PC will have a 3D accelerator in it. It is expected that over 78 thousand accelerator chips will be sold in 1999, with a market value of U.S. $1.1 billion. The Sony Playstation is selling at a rate of 20,000-plus units a year and the industry cannot wait until the next generation Sony home video platform arrives in 2000. Yet, behind this façade of success lie critical issues that keep “Z” from achieving its true potential. We will explore these issues and outline what is required to make 3D a truly ubiquitous technology.

As trite as it may seem, 3D is only about adding a Z coordinate to the imaging of objects or abstract entities. The key value of this is that doing so gives the observer a variable viewpoint and allows interaction with the contents of the scene. Thus, one’s viewing direction is not fixed by the vantage point defined by those who created the content. When doing visualization, this flexibility is all about the freedom to examine. Another component is time. That is, animation is really “past time 3D,” where what one sees is determined by the content creator; whilst “real time 3D” affords the participant more flexibility (with the attendant quality tradeoffs). However, these points do not address the critical issues of adoption and use. Consider the following statements:

  • The success of 3D technology will be determined when millions of individuals can create content and usefully interact with it. The criteria are the same as those that the PC, the home video camera, film and digital cameras, the cell phone and email fulfilled. This is also about casuality; that is, the ability of the user or participant to define the content; what is to be done with it and how; and the interaction anytime anywhere.
  • Real time 3D is the only time-performance element that is meaningful. As good as 3D special effects are, they are still past time. Animation programs on the PC have had only minimal success as mass-market software. Thus, another important criterion is that the creator or participants not know they are dealing with anything less than a continuous representation. From a visual standpoint this implies 60Hz update rates - the same as high-end flight simulators.
  • The other major factors at play are quality and scene content. These also set the complexity of the underlying technology. We will defer discussion on this topic until later in the article.
Table 1: 3D & Data Sources

Examine the Data

In order to address what to do with 3D we will begin by examining differing types of data. Our perspective is very broad as shown in table 1. We seek to show most of the types of objects or data sources in the world along with their attributes. We begin in the first row with a physical space that describes the world, and then we move down the table to a 2D representation in the last row. The telling information in the table lies along the top, with the columns. Here we describe the form of the data; its category (man-made or natural); some examples; availability in digital form; whether a time component is useful; the relationships of the data to video, and lastly, both the relationship with 3D, and its current use in 3D. The key conclusions we can draw from this table include the following:

  • The lack of readily available data is a major barrier to entry. For example, if one needs a model of a city, such as New York, it is just too expensive to gather. Furthermore, human organs have only recently become available in high quality 3D form through the Visible Human project. Many man-made objects exist in digital form, in terms of CAD data, but this is typically neither readily available, nor in a format useful for imaging. In summary, we live in a world in which we are surrounded by man-made and natural objects, but few exist in an available and useful digital form.
  • The entry under data type - Abstract Data Representations - covers virtually any form of data that might exist in a corporate database on in a research lab. 3D can have a role here as a visualization tool. Certainly 3D is already highly utilized in areas such as Computational Fluid Dynamics (CFD), but the applications are much broader than this. Atoms are another form of data that can be readily visualized.
  • The last column, Current State of 3D Use, is a reflection of both the barriers and opportunity.
  • This table is just a way of exhibiting the world about us, be it in natural, man-made or abstract representation. However, it is clear that 3D is a tool utilized in only a very small fraction of that world of data. Thus, the lack of data useful in 3D imaging is both a barrier yet also a major opportunity.

In examining the data in the table, another advantage of 3D is omitted - its infinite zoom ratio. With access to complete data sets, it is possible to zoom from a galaxy to the solar system to the Earth to a city to an object and to its atoms. No other technology, including photography, allows such continuity, and the ability to stop at any point in the zoom and explore the data.

Table 2: 3D & Application

Quality and Content

Now that we have examined the content our attention will be turned to quality issues, and indirectly what can be done with that content. These issues are represented in Table 2. Our objective here is to relate the 3D environment to be created, with the types of data used and the potential of the market.

We see only four classes of 3D environment:

  • Real emulation
  • Recognizable
  • Artificial
  • A combination of a real world with any of the above

This may seem as an overly narrow classification, but we see no other options. In the case of real emulation and recognizable, these form criteria for quality of the space. In the case of real emulation, this mandates that the objects seen by the observer look sufficiently like the real object that they are obvious to comprehend. The criteria can apply to cars, virtually any object and to individuals. The gap between the physical object and the 3D representation, at the resolution seen by the observer, is immaterial.

Recognizable objects do not necessarily have to be those seen or experienced in everyday life. For example, if one sees an R2-D2 from a Star Wars movie this can be recognizable, but few have actually seen a “real” one! Recognizable can also be a meter on quality. Real emulation would mean that the most devout car fan could recognize a model of a 1999 Viper, but recognizable may imply that a sports car could be differentiated from a sports utility vehicle. In gaming most of the content falls into the recognizable category.

Artificial environments span content including representations of abstract or financial data. This is an area of considerable potential, but one which has remained largely undeveloped. As can be seen from the table, this class of environment significantly shifts the traditional 3D barriers to entry. That is, as we have discussed earlier, the major barrier to creating environments is the lack of 3D data and thus models. Here in the artificial class there is typically an abundance of data, and often it exists in more than three dimensions. Thus, the problem of creating useful 3D applications shifts from the objects themselves to the process of deciding just what is useful to display.

In some respects this is a middleware issue, where the software complexity lies between the data sets and the environment representation. Data mining is an excellent example of this complexity issue. The artificial environment class is one that illustrates the issue – “to Z or not to Z.” 3D only provides Z, and to make Z useful, the representation of the data must be in a form which can be depth-visualized in order to enable greater insight, or “add value.” This seems rather obvious, but there are only a few products in this space, and their success is not yet evident. Barriers to making 3D useful have proven to be very elusive to eliminate. Just when the data source problem is eliminated, as in the case of artificial data, the barrier rises yet again, but this time as a middleware issue.

The combination of both real and 3D environments is a special case. A good example would be virtual sets for television productions. Note that here the quality criterion is one where the 3D quality should be the same as video. As appealing as this market may be, it is quite small and certainly not mainstream. Another facet of this same space is augmented reality, where 3D objects are superimposed upon a real environment - the inverse of a virtual set. The implementation usually requires special visual glasses or some other transparent display technology, and this is also a narrow market (especially the case when the glasses are high cost).

Implicit in the discussion of the second table is quality. We believe that the market driving quality criteria will be video. That is, when 3D creates images of the quality which can be integrated seamlessly with today’s video, there will be no explaining required of what 3D is trying to show. Yet, the technology remains distant from this goal, and in many respects what is likely to happen is a convergence of the quality and the barriers to entry at the same time - both being long term. As the quality approaches video we expect that the tools for creating high quality 3D data will be more readily available, and that data sets will likewise be more available.

Looking Back

I was fortunate to see some of the technology and equipment developed at the University of Utah by Ivan Sutherland in the early 1970s. His vision of what 3D could do was very evident in this pioneering work. Yet, in many respects the barriers to making 3D truly useful remain to be overcome. 3D that satisfies many of the criteria described in this article is only being fulfilled in flight simulators. This was the also the case 10 or more years ago. In spite of the fact that the installed base of 3D accelerators is above 100 million units, the same problems exists as did then - how to make it easy for millions to develop and use 3D applications?

3D is both a fundamental expression of computing, and the ultimate user interface. Yet, in spite of 30-plus years of research, the same basic problems exist today that impeded large-scale deployment and use then. The question faced by both the research and development community remains - to Z or not to Z?

Scott S.Fisher is a Media Artist, Producer and Director whose work concentrates primarily on immersive environments and technologies of presence. Currently, he is President of Telepresence Media, a company focusing on the art and design of virtual environment and remote presence experiences. He is also on the Faculty of Environmental Information at Keio University in Fufusawa, Japan.

Glen D.Fraser is a Computer Engineer with a passion for virtual reality and other forms of real-time visual computing. He currently works at SOFTIMAGE, developing interactive viewing and animation tools.

John Latta
4th Wave, Inc.
Alexandria, VA

The copyright of articles and images printed remains with the author unless otherwise indicated.