Oh that is fun, I was one of the operators at the USC Image Processing Institute who did scanning, ran the systems (initially a the HP2100, then a PDP 11/40, and later a PDP 11/55t. We did lots of tours and we had a number of graduate students working on various issues. One of the things we did was to provide the 512 x 512 set of reference images on 1600BPI magnetic tape to researchers. Each image was stored as a red image, green image, and blue image (8 bits per pixel for 24 bits total). In addition to Lena there were other standard images that involved repeating texture, lots of lines and edges, and variable frequency details. By using the standard images, two algorithms could be compared using their output on the same standard image. This helped researchers in the field compare their work with other published studies.
> Please note that we no longer distribute the following images that were previously available in our database: 4.2.04 (lena), 4.2.02 (tiffany), elaine.512, numbers.512 and testpat.1k. Although these images have played a significant role in the history of image processing, they no longer represent the best examples for future research.
It there a reason why these images "no longer represent the best examples for future research"?
How is Lena inferior to the other "female" images? Lena seems to offer a good mix of highly detailed textures, smooth gradients and sharp edges. Maybe there are copyright reasons but Playboy seems to be fine with the use that the Lena image is made of.
Lena is problematic for technical reasons, it’s not PC pushback. The scanning technology of the time was somewhat limited, and the exact parameters are not known if I recall correctly. And of course it is fairly low resolution by current standards. It’s a scan of a reproduction of a production image on magazine paper - something quite different than a photograph. All of these processes introduce characteristic degradations of the image.
I seem to recall someone approaching playboy to see of the negatives where available to make a new “standard” version but I don’t know of that got anywhere. At any rate, there are several subtly different versions around, which has caused some issues in the past.
Lena originally became popular not because people liked looking at a pretty face (though I don’t imagine that hurt) but because it has quite a bit of variation in texture, with strong and weak edges, etc. Particularly for image compression research, it was a real challenge in the 80s and 90s.
All of the technical challenges are better represented these days by other image sets, that don’t have the drawbacks I mentioned. And of course it is much easier to produce your own sets now, and evaluate performance over a significant number of images...
Low resolution was not part of the motivations presented. The main one, IMO, were that it is highly processed and that there were a few slightly different copies going around making ambiguous which one did people actually use.
Also I would personally add that the level of exposure received could by itself warrant removal from a bigger set, since it would still be used as an example anyway and could skew informal perception.
