Graphics & Visualization: Selected papers

The paper presents an original solution to the fundamental problem of recovering a Riemannian manifold of unknown dimension and topology from a set of points known to belong to it. This work has practical implications in disciplines like image processing, computer vision and graphics because many of their practical problems can naturally be reformulated in these more general terms. The originality of the proposed method is that it is based on a special anisotropic partial differential equation, which acts on an implicit representation of the known set of points, and connects them together in a sound way from the viewpoint of differential geometry. This makes it a solution of choice for modeling complex systems which have no known explicit description but from which large amounts of data are available. One important example is the modelization of high-dimensional image features sets for the identification of image semantic cues or categories and its direct application to content-based image retrieval from large visual databases. This paper won the Best Paper Award at the 7th European Conference on Computer Vision 2002.

The significant processing power available in today's commodity processors and graphics adapters has continued to drive the interest in using clusters of workstations for parallel rendering and visualization. The work presented in this paper allows applications to more easily utilize the aggregate rendering power of a cluster, without imposing a specific scalability algorithm that may not meet the application’s needs. The authors have built a framework, Chromium, that provides a generic mechanism for manipulating streams of OpenGL graphics API commands. Chromium’s stream processors can be extended programmatically, which allows users to solve more general problems than just scalability, such as integration with an existing user interface, stylized drawing, or application debugging. This extensibility is one of Chromium’s key strengths. Chromium allows existing applications to take advantage of the rendering capability of a cluster, without modifying, or even recompiling, the source code. Real-world examples of applications that use Chromium and achieve good scalability are given. Chromium's compatibility with existing applications will likely accelerate the adoption of rendering clusters and high resolution displays, encouraging the development of new applications that exploit resolution and parallelism. This paper was accepted and presented at the 2002 ACM SIGGRAPH conference, the most prestigious graphics conference in the world.

Cutting and pasting to combine different elements into a common structure are widely used operations that have been successfully adapted to other media types such as text and images. Surface design could also benefit from the availability of a general, robust, and efficient cut-and-paste tool, especially during the initial stages of design when a large space of alternatives needs to be explored. In this paper, the authors describe a set of algorithms based on multiresolution subdivision surfaces that perform at interactive rates and enable intuitive cut-and-paste operations. Such operations are valuable in a number of scenarios difficult to perform using existing technology. For example, when designing automobile body parts it is common to work in parallel on a digital mock-up and on a clay model. Using the cut-and-paste technique, a designer can scan a feature modeled in clay and paste it onto a digitally-created object. Other applications include importing shapes from libraries, or combining design components from different projects. This paper was accepted and presented at the 2002 ACM SIGGRAPH conference, the most prestigious graphics conference in the world.

Texture mapping is a common operation used in computer graphics to increase the realism of three-dimensional models at low cost. This work shows how the effectiveness of this operation can be greatly increased by reorganizing the content of the image used to texture map the 3D model. A warping function is computed to distort the image such that areas containing perceptually significant details are allotted more texture real estate at the expense of other less important regions. The warped (optimized) image can then be reduced with less loss of detail than an unoptimized one. The authors show that the size of some common texture images can be reduced by up to 80% with unnoticeable loss in quality. The unwarping of the image is performed automatically through the texture mapping operation. Therefore space-optimized texture maps can be included at no cost in the current graphics pipeline architecture. The direct benefit of this work is an economy in texture memory usage, and in the graphics card bus bandwidth requirements. Any application that takes advantage of texture mapping will therefore benefit greatly from this technology. This work won the Gunther Enderle award for best technical paper at Eurographics 2002.