| | |

Interactive 3D Displays go Big!

NRC-IIT researchers have developed a technique to display very large sensor based 3D models using a desktop PC. With international partners from the heritage and museum sector, including the Centre de recherche et de restauration des musées de France (C2RMF) and the University of Lecce in Italy, the NRC-IIT team has collected data from paintings, sculptures and heritage sites to push 3D visualization technology a step further. These items were ideal for testing and developing the technology because of their complex shape, color and texture compositions.

The GoLD (for Geomorphing of levels of Detail) method was presented at the 2005 SIGGRAPH Conference in a paper called “GoLD: Interactive Display of Huge Colored and Textured Models”. The paper was one of 98 selected from a pool of 461 submissions at SIGGRAPH, the world’s leading professional society for computer graphics and interactive techniques .

"The SIGGRAPH Papers program continues to define excellence in research in computer graphics and interactive techniques. It has long been the finest international forum for disseminating groundbreaking, provocative, and important new work. Our selection criteria follow the highest standards, are very rigorous and only accept outstanding innovations in our field.”
Stated Markus Gross, SIGGRAPH 2005 Papers Chair

NRC-IIT Discovers GoLD

Since 2002, researchers at NRC-IIT have been working on a project to improve tools used for the display of 2D and 3D data. They have developed a technique for view-adaptive multi-resolution display called GoLD. 3D sensors and digital cameras allow us acquire very large amounts of data from real objects and/or environments. Data from a single image or model can fill a whole hard disk, thus making the analysis, processing, and rendering of such data into 3D displays heavily time consuming.

This innovation is of significant importance to the development of interactive 3D displays and visualization tools: This project was started in the context of a virtual heritage collaboration with the University of Lecce, where at that time the only way to interactively display the detailed models produced was to discard up to 95% of the data collected, sacrificing a lot of precious information. With GoLD, application specialists can navigate the full detail of huge 3D datasets and therefore understand and exploit it to its full value. Not only have NRC-IIT researchers found a way to utilize all data collected, but their technology provides several major advantages including:

High performance rendering – The method makes full use of the new advanced rendering hardware developed for the gaming industry that is now available in desktop PCs.
Absence of visual artifacts in model – With geomorphing( progressive geometry/color data interpolation), smooth appearance can be obtained at lower geometric resolutions.
Application-adaptive pre-processing – The processing can be optimized for different types of models, especially those with large quantities of high resolution textures such as from mapping digital photographs, a current trend in 3D modeling.
Access to original model at full resolution – Even if the data is transformed to produce the high performance rendering, users can still look at the untouched original when zooming on details of the model.
Limited CPU usage – Most real-time computations are done on the graphic hardware, leaving more CPU resources available for other tasks.

The most compelling achievement from this technology is that 3D rendering and display of those huge datasets can be done using a desktop PC or a laptop. For displaying such models on a PC, the computer needs to produce in real time many images every second. For each second seen on a monitor, minutes of processing and rendering time are required using traditional display techniques. The solution lies in the fact that even if we have billions of 2D and 3D samples to display, a basic monitor only has around a million pixels, so the challenge is in being able to rapidly select and send just the right data that will produce a perfect image on the screen. With the GoLD method, all the data collected is transformed to allow a high quality, accurate display of the actual object much faster and more effectively than with traditional techniques.

This set of features provides a large application scope for this technique, as it can be used wherever data acquired by sensors needs to be displayed. As technology and visualization tools evolve, many industries are taking advantage of these tools to help them in analysis, modeling, quality assurance, preservation and manufacturing stages in research and product development. Aerospace, automotive, transportation, heritage and archaeology, defense and manufacturing sectors are key markets for this new technology, representing billions of dollars worldwide.

NRC-IIT researchers are continuously working on improving the technology as it applies to the entire visualization pipeline (acquisition of data, processing, rendering and display). Through ongoing collaborations with heritage (C2RMF, CMC, Lecce U.), space (CSA, NASA) and industry partners (Neptec, Terrapoint), the team strives to develop more automated and more scalable solutions that will extend the application field of 3D sensor technology while reducing the costs incurred by man hours to process and filter the data. This will make GoLD more appealing and applicable to more markets.

For more information, please refer to the information available on the NRC-IIT website at:
http://iit-iti.nrc-cnrc.gc.ca/publications/nrc-48126_e.html.

Contact

Louis Borgeat
Research Officer
Visual Information Technology
NRC Institute for Information Technology
1200 Montreal Road
Building M-50, Room 353
Ottawa, ON K1A 0R6
Telephone: +1 (613) 991-5477
Fax: +1 (613) 952-0215
E-mail: Louis.Borgeat@nrc-cnrc.gc.ca