December 27, 2004, Updated September 14, 2012

Hebrew University whizzes have developed a mathematical algorithm that requires marking images with only a few crude squiggles to create a highly realistic color picture.Israeli brainpower provides much added value to computer software, according to Science magazine, which recently ranked Israeli computer science as the best in the world. And using the same hardware as anyone else, researchers at the Hebrew University of Jerusalem’s Benin School of Engineering and Computer Science have forged innovations unattained elsewhere.

For example, instead of painstaking manual colorization of old black-and-white movies, the Hebrew University whizzes have developed a mathematical algorithm that requires marking images with only a few crude squiggles that in about 10 minutes creates a highly realistic color picture. Another team imitates the natural abilities of the human eye, optic nerve and the brain to depict gradations of light and shadow by using a mathematical algorithm to meld images of various exposures.

A team of computer and engineering experts at the school are responsible for the world’s first software to allow the creation of panoramic 3-D photographic images of up to 360 degrees with a single digital camera; these can be printed in 2-D on an ordinary home printer and glued to a cheap, lenticular plastic sheet to create an in-depth, hologram-like image. The success of their efforts can be viewed in and on international airports, bus shelters, train stations, phone kiosks, vending machines, magazine covers, brochures, inserts, mailers, CD covers and packages around the world.

The Hebrew University?s mathematics department was founded at the launching of the university in 1925; it took 44 more years until the computer department was started. Originally, university management had decided to leave engineering to the Technion-Israel Institute of Technology in Haifa, but in 1999 it changed its mind, after realizing how math, computers and engineering had become inseparable. The aims are to shape an industry in which changes are endemic, and prepare the next generation of hi-tech pioneers with a dynamic, multi-disciplinary academic program integrating theory and application.

Now celebrating its fifth anniversary, the school has 39 faculty members and 840 students pursuing bachelor’s, master’s and doctoral degrees. Among the honors received by its faculty are the Turing Award and the Nevanlinna Award. The curriculum includes strategic planning courses, and students may study strategy for launching start-up companies. They also participate in joint projects under the guidance of faculty and industrialists, where they learn to provide practical solutions for real problems. Fortunately for the capital, the majority of its graduates work in Jerusalem hi-tech companies such as Intel and NDS.

At a study day for science reporters organized by ISRAEL21c, leading researchers in the school presented their wares. Introduced by senior lecturer Dr. Yair Weiss, the faculty presented some of its active research groups, which include machine learning, computer vision, computer-aided surgery and medical image processing, artificial intelligence, computational biology and computational neuroscience.

Dr. Dani Lischinski, a senior lecturer at the school who runs the computer graphics lab, tempted his guests with grayscale photos of two actresses – Ingrid Bergman and Lucille Ball. With growing interest in old films, Hollywood has sought colorization, but the process is very expensive. The movie Casablanca was transformed into color in a very expensive and complicated manual process some years ago, but the result left much to be desired.

“There are no good automatic tools do this. It requires a lot of trial and error,” Lischinski explained. But to fill this niche, his team developed a patent-protected semi-automatic technique to colorize black-and-white prints or videos.” One merely has to scribble approximate colors in a handful of regions in each picture, and the computer program fills in the spaces with color gradients to make the image very lifelike and authentic. Ingrid looks sexy in a mauve wool sweater, and Lucy’s flaming red hair suddenly appears. It looks like magic.
“We have received a lot of interest from Hollywood studios,” says Lischinski. “We must be doing something right.”

In another presentation, Lischinski displays photographs taken at the Beit Belgia guest house on the nearby HU Givat Ram campus. The image was photographed from the lobby looking toward the sunny courtyard beyond glass doors. But because the lobby is relatively dark and the outdoors very bright, no one camera exposure can depict the scene as the eye sees it; any attempt loses much of the texture and detail. The marble floors looked as if they had leopard spots, and the garden was completely bleached out.

The team coded the light intensities of certain colors, with the goal of compressing the dynamic range. The images are then “squeezed” to avoid overexposure and underexposure, and the result is displayable on an ordinary monitor. He compared an indoor image of a stained-glass window taken in a Washington cathedral: if the colors of the stained glass come out distinctly, the architecture of the window is a blur. Images from teams who worked on this abroad were substandard, but that produced by the HU algorithm technology was nearly perfect, resembling what the eye sees naturally.

But the most eye-catching presentation was that of Prof. Shmuel Peleg, a HU and University of Maryland graduate whose recent research interests include image motion analysis and “image mosaicking.” He, Gideon Ben-Zvi and colleagues took his patented technique of combining 3-D stereo and panorama – which was previously considered impossible – and in 2000 established HumanEyes Technologies. The company’s HumanEyes3-D software was the first to allow the creation of natural panoramic 3-D images of up to 360 degrees with a single digital camera held at arm’s length, and the reuse of the same shot for multiple applications for print or display on 3-D monitors. Four years after its founding, the company remains the only one in its field.

Some of the brands already generating 3-D or motion campaigns with the software include the Miss Universe Organization, Looney Tunes, Coca Cola, Tati, Castro, Coffee Bean, Segafredo and Dove Shampoo. With offices in Jerusalem and the US, HumanEyes now employs 30 talented and experienced staffers.
Until now, anyone who wanted to take a 3-D photograph needed many cameras in different positions to snap pictures of a scene from various angles. All these views were then merged to create a stereoscopic image, which means that, just as in real life when we look at something, each eye sees a slightly different image, and the brain combines the two to create a three-dimensional picture without having to wear 3-D glasses, Peleg explained. But HumanEyes does away with all the cameras. The HU-developed software, Peleg said, creates a stereoscopic photograph by taking the pictures from a single digital camera that has been swept across a scene in a continuous mode. The computer algorithm selects strips from a series of images, each selected from different angles, and then puts a composite image together to produce depth and uncover objects located behind others. Alternately, a number of still pictures of the scene can be used. It is very different from ordinary magnification or zoom.

“We believe it will do to two-dimensional photography what color did to black-and-white photography – that people will come to expect three-dimensional pictures,” Peleg added.

The company is now “trying to persuade” two giant computer printer companies, HP and Epson, to make printers that will print 3-D images directly on the special plastic, thus eliminating the need to glue the printed paper image to the lenticular material. If all goes well, he said, such printers could become available in a matter of months.
Keep your eye on HU’s Benin School… It has many more “magic” tricks up its sleeve.

(Reprinted with permission from The Jerusalem Post)

More on Innovation