For example, the colors and registration will be different than the original image, so it's not like a new scan will identically include the same pixels as the old scan. If a new scan is from the paper copy, then there might have been a change in the ink colors over time. If the new scan is from the negative then there's a difference between that color and what was printed.
You claim that "humans are hardwired to recognize nudity." Is that really true or is it a hypthosis? How was this tested? What is the signal difference between fully nude and partially nude? In the Lenna case, she is not completely nude - how will that confound the results? Does wearing a bikini affect the results? What about gloves? A cummerbund?
I grew up in a white-dominated culture and have more experience with white skin. I've noticed that it's easier for me to identify white skin than black skin. This suggests that recognition has a strong learned component. In those tests, how was the influence of that learned behavior disentangled?
Is the 'hardwired' recognition due specifically to nudity or is it to the transgression of cultural norms? That is, some cultures de-emphasize nudity as being something other than just a person without clothes. This is not generally true of US culture. I can conjecture that there is a hard-wired system to see if a person is following cultural norms, so in the US, nudity would stick out more than in cultures where casual nudity is more common.
Now that I think about it, any hardwired signal like this should be easy to test. Have an array of, say, 10x10 photographs, where one is nude and the others are wearing appropriately skin-tone clothes (fully dressed, partially dressed, etc). Then time how long it takes to find the nude person. This test can be done across multiple cultures pretty easily, so long as nude images are tolerated.
This is similar to the studies I've read about done on facial expressions, to show that some are very cross-cultural and thus likely more hard-wired, while others are not and thus more likely to be learned. For example, as your baseline reference, use a similar 10x10 grid but where one person is frowning while the others have a neutral face, and time how long it takes to find the frowning person.
Since it sounds quite simple to carry out, I figure it's been done, so if that's the sort of work you're referring to, can you kindly point it out? A Google search gets bogged down with references to algorithm for detecting nudity in images.
Anyways, you're absolutely right, people's brains are very very very much optimized for evaluating the microfeatures in a person's appearance, so its actually the right perceptual way to benchmark lossy compression (or at least one very good way).
http://www.ebay.com/sch/i.html?_trksid=p2050601.m570.l1313.T...
Unless they got the negative from playboy, I don't think it's an appropriate image to use anymore.
Or is it just the fact that photography back then was done with negatives, and anything but the original photo is unsuitable for compression research?
"The Muirhead had a fixed resolution of 100 lines per inch and the engineers wanted a 512 x 512 image, so they limited the scan to the top 5.12 inches of the picture, effectively cropping it at the subject's shoulder."
It's not about wanting to see nudity. The original is useful because humans are hardwired to recognize nudity. It's not a coincidence that tasteful nude pictures are considered fine art. There are literally a hundred thousand tiny features that are lost by cropping the image to her face and shoulders. Those features are much easier to compare than, say, pictures of nature. Artifacts introduced into pictures of leaves and grass are far less noticeable than artifacts introduced in the subtle variation of the colors across her back.