[ Up ] [ Graphics ] [ Digital Libraries ]

New Opportunities and Challenges for Computer Graphics

The proposed Intel Architecture platforms offer new opportunities for developing distributed graphics-intensive applications that leverage higher processor speeds and specialized accelerators for video and graphics performance. Current workstation graphics algorithms will need to be re-evaluated to take advantage of different hardware configurations and resources. New ideas such as image-based rendering show potential for increased performance and open the doors to new applications in telepresence, but their effectiveness as alternatives to traditional geometric modeling and object-based rendering has not been fully demonstrated. Foundational research will be required to prepare the next generation of rendering algorithms to best utilize the next generation of Intel hardware.

The Program of Computer Graphics has served a primary role in the development of realistic image synthesis since its inception under the leadership of Donald Greenberg in 1974. As we enter our 25^th academic year of operation, we continue to emphasize the need for foundational research in graphics algorithms. Our research in global illumination spans the full range from physically-based lighting models, light transport algorithms, and perceptual models of human vision to permit greater efficiency in rendering only what we need to for our intended purpose. Many of our algorithms have been adopted in commercial graphics applications, and our research has influenced the development of graphics hardware.

Our long-term goal is to develop physically based lighting models and perceptually based rendering procedures for computer graphics that will produce synthetic images visually and measurably indistinguishable from real-world images. Fidelity of the physical simulation is of primary concern.

In computer graphics we are only beginning to take advantage of our knowledge about visual perception for realistic image synthesis. A better understanding of the spatial, temporal, chromatic, and three-dimensional properties of vision can lead to even more realistic and more efficient graphics algorithms. An accurate visual model allows us to characterize the visual states of the scene and its observers and allows us to relate them to determine the mapping from simulated scene radiances to display radiances necessary to produce a perceptual match between the scene and the displayed image. This allows the images to be used quantitatively in areas such as illumination engineering, safety design, and visual ergonomics.

We also see exciting possibilities in several new image-based rendering methods, essentially high-dimensional interpolation; from a limited set of original captured images, or a limited set of synthetic rendered simulations, further views can be calculated from different viewpoints. These techniques offer particular promise for applications in telepresence, where the full inversion to reconstruct scene geometry is not required, and the goal is merely to render a sufficiently accurate view of an environment from any desired vantage point.

Our proposed research will investigate the limitations of the more promising of these techniques. With our experience in global illumination, we will seek to identify what additional information about the lighting or a priori knowledge of scene geometry can most readily improve the results of these techniques. We will pay particular attention to possibilities for hybrid approaches combining elements of object-based as well as image-based rendering, given the hardware resources of the Intel platform.

Participants

Web Links

Last modified on: 07/30/99