New software processes hundred-billion-pixel photos in seconds
By Paul Ridden
October 31, 2010
Computer scientists from the University of Utah have developed computer software that allows editing of "extreme resolution" image files in a matter of seconds, a process that could previously have taken hours. Whereas existing editing suites require the full gigapixel image to be loaded into a computer's memory before manipulation can begin, the new development draws a lower resolution preview image from an externally-stored image into the editing screen. Users are said to benefit from being able to make image-wide modifications in seconds rather than hours and on devices normally not nearly powerful enough for such things.
When editing photographs containing billions to hundreds of billions of pixels, existing technologies require full resolution images to be loaded into an application before editing can begin. This, in turn, requires enough memory and processing power on the computer equipment to handle all that data without seizing up. The Visualization Streams for Ultimate Scalability, or ViSUS, software does things a little differently.
ViSUS works by using a sophisticated algorithm to choose the required subset of available pixels to produce an approximate preview of the fully processed monster image which allows editors to work in a fraction of the time taken with existing systems. Valerio Pascucci, an associate professor of computer science at the University of Utah's Scientific Computing and Imaging Institute (SCI), likens the process to public opinion polling where "you ask a few people and get the answer as if you asked everyone. It's exactly the same thing."
During a study of the system, the team loaded in images varying from a few megapixels to a few hundred gigapixels "to test how well the ViSUS software let them interactively edit large images, and to show how well the software can handle images of various sizes, from small to extremely large." In one example, they took a 3.7 gigapixel image of the Earth and spliced in a 116 gigapixel photo of the city of Atlanta, placing it underwater just off the Gulf of Mexico - recreating the lost city of Atlantis.
Pascucci said that the demonstration was "just a way to demonstrate how an artist can manipulate a huge amount of data in an image without being encumbered by the file size."
The team – which is made up of Pascucci, SCI colleagues Brian Summa, Giorgio Scorzelli and Peer-Timo Bremer from the Lawrence Livermore National Laboratory in California – also placed a camera on the roof of a University of Utah building and snapped a 611-image, 3.27 gigapixel panoramic view of the entire Salt Lake Valley. At full resolution, it took some four hours to stitch all of the images together but the ViSUS software offered a workable preview in a matter of seconds.
But, according to Pascucci, the quick preview is only a small part of the power of the new software, its real significance lies with the ability to allow editors to zoom into any part of the low resolution preview for selective editing. If an editor needs to work on a particular segment of the image, ViSUS can zoom into that section and grab additional image data from the externally-stored original. The team says that the software could approximate a terabyte of full resolution image data using just a megabyte of the total image data.
This also means that computers, laptops and even smartphones could use the software to preview, edit and analyze images stored on external drives or servers. "In our method, the preview has constant size, so it can always fit in memory, even if the fine-resolution data keep growing," Pascucci said. Users have the option to leave the original image data untouched, to save the edited changes permanently or to save as a separate file.
The researchers see the development being of use to doctors, intelligence analysts, photographers, artists, engineers and anyone who has gone through at least 20 cups of coffee while the photo editor gets to grips with the huge image being loaded in. It could also be of interest to gamers, who may be able to build custom gaming environments on the fly. Although the researchers have not as yet processed any 3D photographs, they say that too is possible.
The research will be published online shortly in the Association for Computing Machinery's computer graphics journal, ACM Transactions on Graphics. The team now plans to develop the software for commercialization.