Computer graphics and visualization exploit vision, the richest of human senses, to improve communication from computer systems to users. IBM Research's activities in graphics and visualization span the entire range of the field, from data acquisition, representation, and transmission to rendering on a range of hardware from palmtops to supercomputers, with applications to engineering and science, business, and finance.

Acquiring Visualization Data

Different aspects of the problems of high-resolution imaging are being studied . The Image Library Applications project develops technologies for capturing and reproducing high-quality still images. The project focuses on image watermarking and the accurate capture and preservation of image color. Our research has been informed by developing image capture and preparation systems for the Vatican Library, the State Hermitage Museum, and Tiffany and Co.

The Pieta` project concentrated on a different aspect of data acquisition: scanning three-dimensional physical objects at high resolution and fidelity. Our goal is to develop fast, practical, low-cost scanners for capturing the geometry and colors of complex objects. As a test case, we scanned Michelangelo's Florentine Pieta`, a 2.25-meter marble statue, to produce
a photo-realistic computer model accurate to less than a millimeter.

Data Representations for Visualization

Improving a visualization system depends on developing the best data representation for the problem at hand. Whether the goal is to create images faster, to render at higher resolution, or to enhance a user's comprehension of his data, the key is to understand how the data must be
organized to adapt it to the user's needs.

At Tokyo Research Lab, we focus on developing new algorithms that enhance the ability of designers to construct smoothly curved surfaces for mechanical CAD, e-commerce, games, and animation. We are developing powerful new techniques for sculpting virtual shapes and representing them with parametric surfaces.

Today's networked environment requires efficient transmission of data. Traditional three-dimensional data formats are bulky. IBM's algorithms for compressing three-dimensional data 50:1 or even 100:1, now part of the MPEG4 standard, have enabled the development of otherwise infeasible three-dimensional applications.

Exploratory Visualization

In many domains, particularly science and business in which data are often abstract, it is difficult to predict what visualization will convey the most information to a user. In many cases, no single view will suffice,
but a combination of techniques may reveal important facts. Teams researching these problems have developed technologies like OpenDX and Diamond for increasing users' access to their data by providing a wide variety of flexible tools and interfaces. Our research combines the techniques of different disciplines to develop new understandings of the strengths and limitations of computer visualization.

The HotMedia TV project is pursuing another approach to increasing the value of visualization by adding information to images to create interactive video and television.

From Supercomputers to Palmtops

Delivering visualizations to users requires adapting systems to the full range of modern computers. At Watson, we are studying the problems of providing visualizations on the entire range of computing systems.

We participated in the development of a highly parallel graphics system coupled to an IBM RS/6000 SP supercomputer, which can create immersive virtual environments by connecting multiple graphics adapters to a single SP node to share the load of rendering to each of several displays.

Distributed rendering was also the goal of the Adaptive Rendering and Transmission Environment, a system for allowing access to three-dimensional models from a network of varied workstations. It adapts to varying network connections and rendering capabilities.

Bringing the power of visualization to low-end systems opens a new range of challenges. We are investigating the possibilities of pure Java rendering, and our results have been incorporated into IBM's HotMedia technology.

Please contact Paridhi Verma to obtain copies of the Computer Science Brochure