A Nude ‘Playboy’ Photo Has Been a Mainstay in Testing Tech for Decades
The documentary ‘Losing Lena’ is about the many small ways in which women are told they don’t belong in tech
In 1972, a Swedish woman named Lena Söderberg accepted a modeling job from the photographer Dwight Hooker. Söderberg was 21, new to the United States, and broke. The name of Hooker’s employer, Playboy, didn’t mean much to her; the contract definitely did. “It was money, and I didn’t have a lot of money,” she explained to Wired earlier this year.
In the photo shoot’s most famous image, a hat-wearing Söderberg stands nude before a full-length mirror, clutching a feather boa and looking over one shoulder. The photograph ran as the centerfold in the November 1972 issue. Then she moved on with her life.
The following year, a team of engineers at the University of Southern California’s Signal and Image Processing Institute were looking for an image on which to test a new piece of image-compression software. A man in the lab — they were all men — offered his copy of Playboy, because it was the 1970s and bringing a Playboy to work was an okay thing to do.
A colleague ripped out the photograph from the shoulders up and ran the now-PG-rated image of Miss November — or Lena, as the image would come to be known — through a converter. It worked. The lab passed out copies of their successfully compressed image to visitors, which other programmers then used to test their own algorithms and compare with others’ results.
The Lena image was an ideal test case for image-processing algorithms, with rich contrasts, color, and details anchored by the familiar contours of a human face. Other images possessed those same qualities, but this one appealed to the predominantly male image-processing sector. Wavelets, compression, reconstruction, denoising: Whatever the technology, Lena was used to test it. Download one of the many free-use copies of Lena available on the internet, and you get a JPEG — a format developed with the use of Lena.
Even as the technology and the engineers working with it aged and changed, the Lena image did not. It persisted in labs as if it were an inalterable part of the furniture and not the result of an individual choice made in an entirely different era.
“It wasn’t just the issue of the Lena image. It was an image problem that expanded beyond gender.”
It’s the environment surrounding the image, not so much Lena itself, that is the subject of Losing Lena, a short documentary that had its North American premiere last week in an auditorium at the University of California, Los Angeles.
While the picture “was the singular subject” at the beginning of the filmmaking process, as interviews went on, it became clear that “it wasn’t just the issue of the Lena image,” producer Francesca Walker said. “It was an image problem that expanded beyond gender.”
The film was produced by the Australian agencies FINCH and Clemenger BBDO Sydney on behalf of the organizations Creatable and Code Like a Girl, which focus on encouraging girls to participate in STEM fields. The half-hour film premieres on Facebook Watch today. It opens and closes with interviews with the real-life Lena, now a grandmother with close-cropped white hair living in Sodertalje, Sweden.
The bulk of the documentary focuses on the industry that made her image famous. In interviews with academics, educators, and students, the filmmakers examine the ways in which an industry that claims to love disruption clings to its own outdated icons and the people who get excluded as a result.
“Once upon a time, I was the centerfold of Playboy,” says the former model, who now goes by the name Lena Forsen, in the film’s final moments. “But I retired from modeling a long time ago. It’s time I retired from tech, too.”
At the peak of Lena’s popularity, the strongest argument in favor of using the image in research was that so many others had done the same. Stripped from its original context, the Lena image was simply a recognizable pattern of pixels that could be manipulated, compressed, and then compared with the results of other compressions of the same image.
“At the height, it was used in everything from journal papers to textbooks,” said Deanna Needell, a mathematics professor at UCLA. “For a long time, I didn’t attend a single conference in image processing where she didn’t appear in someone’s talk. And now it is still, sadly, not uncommon.” To demonstrate that Lena wasn’t the only available face with the right amount of texture and shading, Needell and a colleague used a photograph of the model Fabio Lanzoni in a 2013 paper on image reconstruction.
The real-life implications of a ’70s-era Playboy centerfold being presented as a neutral image was apparent as recently as 2014. Maddie Zug, then a high school junior, was one of a handful of girls in a mostly male artificial intelligence class told to use the Lena image in a coding class assignment.
Norms originate with choices.
The teacher emphatically instructed the class not to search for the full image on Google, which of course everyone promptly did. Instantly, the awkward experience of being one of few girls in a room of teenage boys became the intensely awkward experience of being one of few girls in a room of teenage boys snorting and laughing over a picture of a naked lady.
“The image is a good hook,” said Zug after the film, in which she’s interviewed. (She wrote a 2015 Washington Post op-ed calling to end Lena use after her school ignored her complaints.) “But the real story here is about diversity and inclusion.”
Losing Lena is about the many small ways in which women (and, though the film doesn’t explore this theme as fully, people of color) are told they do not belong in the tech industry. Gatekeepers often defend their industry’s lack of inclusion as the result of systemic problems beyond individual control — leaky pipelines, cultural fit, women’s own lack of interest in pursuing tech jobs. But norms originate with choices. Surely there were more magazine photographs of Richard Nixon available in 1973 than there were of nude women, but that’s not what the USC engineers reached for. The flimsiness of the argument for keeping Lena make those look awfully shaky, too.
Gender politics aside, the main reason for abandoning Lena is technological: The average smartphone takes photos with exponentially more pixels than a scan of a 1970s-era magazine print. Relying too much on a single test image also runs the risk of creating algorithms that work very well on that one test case and less well on others.
The movement to abandon Lena has already been well underway for the past two decades. Several journals no longer accept papers that use the Lena image.
“In cases where another image will serve your purpose equally well, why not use that other image?” wrote David Munson, editor in chief of the journal IEEE Transactions on Image Processing, in an editor’s note on Lena back in 1996.
“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,” the Atlantic noted in 2016. “In 2015, Lena turned up in only 6.2 percent of articles — some 38 times.” In one very encouraging sign, a search for “Lena” on the journal’s site yields 63 mentions of the name in journal articles from 2017 to today. Almost all belong not to the centerfold, but to the articles’ female authors.
