Designer Bubbles
A Hidden Presence
Designer Bubbles
Inspiration has visited scientists while they were engaged in all sorts of activities, from sitting under an apple tree to dreaming of coiled serpents. Now, an IBM researcher has added washing clothes to the list. Five years ago, while watching soap bubbles form and re-form as a load of laundry sloshed about, Kenji Shimada -- until recently a researcher at Tokyo Research Laboratory (TRL) and currently a lecturer at Carnegie Mellon University in Pittsburgh -- was struck with the idea for a computer-aided design (CAD) technique now known as bubble mesh.
Normally, when engineers want to analyze a CAD-generated object -- say, a car door -- they first generate a mesh that divides the object into tiny geometric elements. Stresses, strains and other physical quantities can then be computed using finite-element techniques, whereby values are defined only at the nodes of the grid, simplifying the calculation.
Generating the mesh is only a partly automated process; a designer must manually replace the standard coarse mesh with a fine mesh around certain critical points -- for example, a window in a door panel -- that require more detailed analysis. This step takes time and introduces the danger that too little gradation between the meshes could mar accuracy.
"Engineers have long been eager for a system to automatically generate a well-shaped mesh with fine gradation," says Masaki Aono, a researcher at TRL. The bubble mesh system resulting from Shimada's insight, says Aono, has both raised the quality of analysis and slashed the time required from one month to one day.
ANOTHER FINE MESH
To create a mesh of virtual bubbles, the design engineer first chooses the radius of the smallest bubbles to be located around critical areas. Then the system automatically interpolates the bubbles with the gradation of sizes needed to fill the remaining space. Because there is no way of predicting how many bubbles will be necessary to fill the space, some areas at first have too many or too few bubbles. To rectify this, the meshing process goes through 100 to 200 iterations, during which bubbles are added or subtracted and are spaced according to the actual repulsion and attraction of physical bubbles.
Once equilibrium has been reached, the bubbles are converted into a conventional -- but optimally spaced -- triangular or quadrilateral mesh. This is done simply by connecting the centers of the bubbles.
According to Takayuki Itoh, another TRL researcher involved with the project, IBM has been working
with a major Japanese manufacturer since 1995 to test the bubble mesh in the field.
After more than two years of testing and customizing the technology to suit its needs,
the manufacturer has decided to implement the system enterprise-wide.
A Hidden Presence
Digital image and audio technologies have made recorded media easier to modify and copy. Along with expanded creative possibilities has come the need for efficient and secure means of protecting copyrights and authenticating digital media.
IBM Research has taken several approaches to these problems. Scientists at the Tokyo Research Laboratory (TRL) have developed watermarking solutions for digital images and motion pictures, as well as for Digital Versatile Disk (DVD) audio. Among the applications of the technology are photo authentication, DVD copy protection and audio watermarking for distribution of music over the Internet.
One technique to foil unauthorized use of digital property relies on embedding a visible watermark on an image. Another makes downloaded images become half transparent. "But we have come up with a key-controlled, visible watermark technique," says TRL researcher Norishige Morimoto. "The watermark is completely reversible. A photographer, say, can distribute samples on a free CD-ROM. Anyone who wants to purchase a picture simply buys the key."
A DAMPER ON FRAUD
Image authentication for the insurance business is another application. Currently, adjusters rely on printed photographs for visual records of damage claims. Digital photos would save money and shorten the claim process -- but they would also heighten the risk of digital forgery. "In the digital domain, there are no negatives," says Morimoto. "A person could forge a photograph without a trace." IBM's solution is to embed into the photograph a special structure that reveals any part of the picture that has been altered.
With support from IBM's Insurance Industry Solutions Unit, TRL entered into a first-of-a-kind project with a major Japanese insurance company. By August, on-site tests had confirmed the technical feasibility of using the IBM techniques for
authenticating photos and detecting
alterations. "If this technology ultimately proves practical, people would be able to trust digital photographs, making
possible an avalanche of applications,"
Morimoto predicts. One valuable use for the technology could be to prevent the alteration of medical X-ray images.
IBM is also ahead of the competition in the much bigger emerging market for DVD video. The company was first approached to look at data hiding techniques for copy protection by anxious motion picture associations in mid-1996. IBM successfully developed a system, and in February 1998, the DVD Data Hiding Subgroup of the Copy Protection Technical Working Group deemed IBM's technology the most capable of surviving a variety of video processing techniques.
In July 1998, IBM announced a collaboration with NEC, in which the two companies would combine their watermark technology to fully address the requirements of the Motion Picture Association of America. The association wants to allow consumers to copy a cable or satellite transmission once, and only once, for their own use.
AUDIO, TOO
TRL is also competing with 10 other organizations to produce an industry standard audio watermarking system for Europe. "The problem," says Morimoto, "is producing an algorithm that can survive not only compression but also the distortion arising from digital-to-analog conversion and cropping during the editing stage -- all without significantly degrading the DVD audio quality."
Such watermarking can be used to prevent illegal copying and to track and identify contents,
just like watermarks for other multimedia contents. "This is a very hot area," says Morimoto,
"and it is quite possible that we will see a digital audio watermarking system in practical
use even before DVD video applications, since it presents fewer standardization and cross-industry
issues."
