Daniel Lemire's blog

, 3 min read

Should computer scientists keep the Lena picture?

There is a very famous picture commonly used in image processing research and classes, that of Lena Forsén. Lena Forsén posed nude for the Playboy magazine in the 1970s. A cropped version of the picture has been used ever since. The picture being used is “safe for work”, but if you Google long enough, you will find the complete version.

I worked for years on the picture, being totally ignorant of the origin story. It is many years later that I learned the story. (I am old enough that I did my Ph.D. in the pre-Google days).

To many women, the use of the Lena picture is offending. This puts me in the awkward position of choosing not to include the picture I refer to in my post. It is available online.

What should we do?

Richard Matthews from the University of Adelaide thinks that we should keep using the picture:

Should the field in general stop using the Lena image? My personal view is: no. The use of the Lena test image is a quirk of the industry that should be celebrated. That being said, it should be used alongside others equally.

I disagree with Matthews.

When I wrote my Ph.D. thesis, I think I was ignorant of the backstory behind Lena, and I used it abundantly in my tests because it is a very rich picture, with lots of different textures. Still, when I wrote my Ph.D. thesis, instead of going with Lena, I used a painting of some famous mathematician because I thought that the theme was more interesting.
There are good alternatives such as the Kodak image set that do not contain any objectionable image.

“Getting along” is how we prosper. There are times when offending people is warranted. But I don’t think that the offense caused by Lena serves a useful purpose.

We cannot remove the Lena picture from the articles and textbooks, and we should not seek to make the picture disappear… but in the interest of “getting along” we should use alternative pictures.

Getting along is a statistical concept. You can never please everyone or even most people. If you remove anything that could offend someone, somewhere, you will be forced to hide in a very dark hole. In the case of the Lena image, in my view, enough reasonable people are offended that it is worth taking it into consideration.

Of course, you can never win cleanly. It is entirely possible that other people would be offended at the thought that we might remove the picture from consideration. Hence, I use the term “statistical”. It is not black and white. Wrong and right.

I will conclude with a personal story. In the lecture notes of one of my courses, there is a picture of a prostitute by the side of the road. If you are not paying attention to the content of the course, it seems totally out of place and maybe sexist. At one point, we were paying a revisor who went through my notes. When she got to this picture, I think she was ready to throw her hands up and call the cops. Except that the cover in question is that of a feminist magazine who was one of the first to benefit from digital archiving. It was, historically, a very important feminist (I stress feminist) magazine. That particular issue had a piece on prostitution. I used this cover because that’s the cover the people who completed the digital conversion promoted; in the original announcement, there were famous feminists being photographed right by the cover in question. I could have removed the cover in question from my notes in the interest of “getting along”, but one should draw a line somewhere. My point is that we should not bend so far as to enter a 1984-like Orwellian world. We should not go so far as to forbid the Lena picture, or to clean the world of anything that might offend. Still, I would probably have chosen a different cover had I thought it could offend (needlessly).

Further reading: Should we discourage the use of Lenna?