Clik here to view.
When a 21-year-old Swede named Lena Söderberg became Playboy magazine’s Miss November in 1972 under the name Lenna Sjööblom, there was little to set her apart from other Playmates of the era. She was a classic beauty, demure yet approachable, but otherwise unremarkable in the pantheon of Playboy centerfolds. It wasn’t until a few years later that Lena, with the help of a few researchers at the University of Southern California, would earn herself an unexpected place in history.
In the mid-70s, with funding from the Advanced Research Projects Agency (now known as DARPA), a team at the university’s Signal and Image Processing Institute was doing pioneering work in the field of digital image processing. As Andrew Sawchuk, then an assistant professor of electrical engineering at USC, recalled for an industry newsletter in 2001, the team members typically tested their algorithms by just scanning whatever stock images they had on hand. But one day in 1973, they happened to be looking for a glossier alternative, something with a human face—the same day that one of their colleagues happened to walk into the lab with Lena’s issue of Playboy. Limited to images that would fit in their 512-by-512-pixel drum scanner, the team ripped the top third of the centerfold out of the magazine, cutting off Lena at the shoulders.
Over time, the researchers built up their own collection of test images—Lena, a colorful mandrill face, and a group of bell peppers, to name a few. Scanners were a rarity at the time, so the USC image library quickly became the standard set of tools for imaging scientists around the country to test their own processing algorithms. From a compositional standpoint, Lena in particular was an ideal image: Feathers in her hat provided great detail, a human face provided a recognizable subject, and a variety of gradations and textures proved to be a useful challenge for processing algorithms.
Unbeknownst to Lena (and, for many years, Playboy), the photo quickly became the single most widely used picture in image-processing research. She was one of the first pictures uploaded to ARPANET, the precursor to today’s Internet. And perhaps most importantly, Lena was used to develop the now-ubiquitous .JPEG image format, a compression scheme that allows complex digital files to transfer between devices and appear on smartphones.
A Google Scholar search for “Lena image” now returns more than 18,000 results. Lena’s featured in authoritative textbooks like Digital Image Processing, and she’s served as a test subject for scanning, autofocus, color-correction, and countless other editing tactics. New work using her likeness is published on a monthly basis, and Lena herself—the person, not the image—made an appearance at the International Conference on Image Processing awards banquet last September. “It was like meeting a celebrity,” says Hoda Rezaee Kaviani, a Ph.D. candidate at McMaster University in Ontario, Canada, who won an award for her video-editing research at the event.
But her ubiquity has also propelled Lena to the center of a polarizing debate within the field. Some view her as an important part of image-processing history, a key brick in the path that’s led researchers to the Internet, cameras, and smartphones of today. Others see another, more-troubling layer to Lena’s symbolism, arguing that a Playboy centerfold—even one cropped to a PG rating—is just one more message to women that they don’t belong in the male-dominated world of computer science.
The field’s gender-diversity problem is well documented. Research from the American Association of University Women in 2015, for example, found that women only account for 8 percent of computer-software degrees and 10 percent of computer-engineering degrees. And among those degree-holders, only 38 percent remain in the field in the four years after graduation.
Some researchers partly attribute these low numbers to subtle, yet consistent, hints of sexism that add up over time. “There are so many micro-aggressions and so many structural barriers that are put up for women and for other minorities in computer science,” says Colleen Lewis, an assistant professor of computer science at Harvey Mudd College. A 2009 University of Washington study, for instance, demonstrated the degree to which classroom surroundings can influence women’s sense of belonging. The experiment placed students in either a stereotypical computer-science lab, replete with Star Trek posters and soda cans, or a lab decorated with nature posters and coffee mugs. Women who entered the Star-Trek environment showed less interest in computer science; men were unaffected.
For women in computer-science classrooms, the Lena image can feed into this so-called imposter syndrome. Deanna Needell, now an assistant professor of mathematical sciences at Claremont McKenna College, recalls the first time Lena came up in one of her graduate lecture courses: “All of a sudden, I was singled out for my gender, which was already sort of uncomfortable, because I was in a room of 60, and there were maybe three women.”
Some imaging scholars believe that the Lena image has become so divorced from its original context that it shouldn’t be held up as an example of sexism within their discipline. “When you use a picture like that for so long, it’s not a person anymore; it’s just pixels,” says Jeff Seideman, a former president of the Society of Imaging Science and Technology.
Needell, though, became exasperated by the Lena image’s constant presence in papers and at conferences—and eventually, she decided to flip things around. For a 2013 image-processing study, Needell and her co-author Rachel Ward obtained the rights to an image of the Italian heartthrob Fabio Lanzoni. The study became something of a flagship for the anti-Lena resistance within the field, and Needell has received many requests—nearly all of them from male researchers—to use the Fabio image in their papers.
On a separate level, Needell hopes her objections to Lena will help her fellow researchers broaden their horizons a bit. “There is a joke now that if you want to write a good image-processing algorithm, you just have it output the Lena image,” Needell says. “Everyone is just fine-tuning their algorithms to this specific image.”
It’s a point that members of the image-processing community have been considering for decades now. In 1996, David C. Munson, then the editor-in-chief of the journal IEEE Transactions on Image Processing, wrote an editorial responding to calls for a Lena ban. Ultimately, he didn’t choose to ban the image, but did caution researchers about its overuse. “We could be fine-tuning our algorithms, our approaches, to this one image,” he says. “They will do great on that one image, but will they do well on anything else?”
Researches today have access to tremendously large datasets, allowing them to test their algorithms on dozens, even hundreds, of images. Even so, Lena is still disproportionally chosen to illustrate those algorithms in the final, published work.
In recent years, though, her influence appears to have waned a bit. In the year Munson penned his editorial, there were 57 mentions of Lena (or “Lenna”) in the journal, accounting for more than 30 percent of its articles. In 2015, Lena turned up in only 6.2 percent of articles—some 38 times.
But for Scott Acton, the current editor of IEEE Transactions, that may still be 38 times too many. In December, Acton and two former editors revived the old call for a ban, and proposed to the journal’s editorial board that IEEE Transactions institute a moratorium on Lena research. “In 2016,” he says, “demonstrating that something works on Lena isn’t really demonstrating that the technology works.”
While Acton does believe that the Lena image “doesn’t send the right message” to female researchers about their inclusion in the field, his strongest objections are technical in nature. Lena contains about 250,000 pixels, some 32 times smaller than a picture snapped with an iPhone 6. And then there’s a quality problem: The most commonly used version of the image is a scan of a printed page. The printing process doesn’t produce a continuous image, but rather a series of dots that trick your eye into seeing continuous tones and colors. Those dots, Acton says, mean that the scanned Lena image isn’t comparable to photos produced by modern digital cameras. Short of an all-out ban in the journal, he says, making authors aware of the image’s technical and ethical issues might be a way to usher Lena gracefully into retirement.
Even Seideman, who’s dubious about the ethical objections to Lena, admits that image processing will likely leave her behind eventually. “It’s a historical footnote,” he says. “If it wasn’t her, it would’ve been someone else. It could’ve been the mandrill.”
Image may be NSFW.Clik here to view.