Some recent projects:

	Efficient Triangle Representation The goals of this project are two-fold. The first is compact representation of triangle data. This includes vertex compression as well as topology compression: in a form that can be efficiently decompressed on the hardware. The additional goal of this work is to arrange the compact representation in a way that on decompression the triangles appear in a cache-friendly order. This implies an order where vertices used by a triangle are likely to have been used by a recent triangle and one could skip the vertex re-shading, simply obtaining the results from a cache.
	Seamless Texture Atlases Texture atlas parameterization provides an effective way to map a variety of color and data attributes from 2D texture domains onto polygonal surface meshes. However, the individual charts of such atlases are typically plagued by noticeable seams. We describe a new type of atlas which is seamless by construction. Our seamless atlas comprises all quadrilateral charts, and permits seamless texturing, as well as per-fragment down-sampling on rendering hardware and polygon simplification. We demonstrate the use of this atlas for capturing appearance attributes and producing seamless renderings.
	Digital Hammurabi Digital Hammurabi is a major, cross-disciplinary effort aimed at making and visualizing highly accurate three dimensional models of cuneiform tablets. This project includes the design and implementation of a capture system capable of scanning small objects at a resolution of 50 micrometers or more. We are developing algorithms for managing, storing and remotely displaying the massive amounts of data generated by the scanning system. We are also developing novel display algorithms to enhance the small tablet features.
	Volume Visualization Our visualization project involves interactive visualization of massive scalar fields. These fields may be defined on structured grids or unstructured grids. In addition, they could be time-varying. So far we have focussed on guaranteed speed visualization of fields too large to fit in the 3d texture and on interactive exploration of iso-surfaced to enable knowledge discovery. We are using some of these techniques to assist physicists to learn about the characteristics of turbulence near black holes.
	Scalable Walkthrough Walkthrough of large out of core models has been an active area of research. Our contribution to this field is an approach that is highly scalable. It ensures high-quality (with a deviation error of less than 1 pixel on screen) display of models of increasing sizes at interactive rates at the cost of some pre-processing.
	Visibility Computation Visibility computation is required for rendering and shadow generation. In order to reduce the bottleneck on pixel processors, we have developed schemes to cull large portions of data, which are guaranteed to be invisible. We have developed algorithms both for back-face culling and occlusion culling.
	Spline Surface Rendering We have developed algorithms for fast view-dependent triangulation and display of spline surfaces. Some of the early work was limited to uniform tessellation. More recent part of the project has focussed on adaptive tessellation and point based rendering of spline surfaces.
	Surface Repair An overwhelmingly large fraction of models in use today have errors, numerical as well as topological. We have developed a system to automatically find errors in a large model and correct them. This system includes a human-in-the-loop step, which allows a human to navigate the space of possible corrections and select one more suitable for the application than the one derived by the system.

See this list of selected publications for more detail.

** Revised: Dec 2006