This is how research progresses. Datasets become more and more difficult. Algorithms become more and more sophisticated. Some breakthroughs are made. And the rest, as the cliche goes, is history. In that sense, the whole field of image processing made a remarkable breakthrough recently. Before I tell you about it, some background …
It was the late ’90’s, a few years B.G. (Before Google became famous). I was an undergraduate student in India in a beautiful seaside campus. The campus came with a private beach – not that most of us needed a tan. Still, a nice place to relax … Being a technical institute located far away from metro cities like Bombay and Bangalore meant that the student demographic was chromosomally skewed – Y numbers dominated overwhelmingly.
It was here that I was took my first lessons in Fourier transforms, Laplace transforms, z-transforms and all things image processing. Oh, do I still get goose bumps thinking about those signals and systems?! (Yes, I have the nerd glasses to prove it!).
During these classes, we came across a standard test image in image processing – the famous lena image. The lena image has been used as a standard to assess performance of various image processing algorithms – blurring, reconstruction, wavelets, and so on – for nearly four decades now.
Naturally we wondered whose picture this was. And how it came to be so popular. But this was B.G., so answers were less forthcoming. It did lead to a lot of speculation during mellow, drunken nights on the beach when we discussed all-important questions surrounding CS vs. EE.
“Oh, stop complaining about your PN junctions. You guys at least have pretty pictures in your textbooks. Just look at your Lena. What do we have? Outdated flowchart diagrams! Sheesh!”, the CS guys would complain.
The more, shall we say, well-read guys in the group were convinced that the image is from one of two sources – two magazines which were not stacked in the library, but very popular in the dorms.
It would be 5 or 6 years more before I realized that they were indeed right. The story of how the Lena image came about is now merely a couple of clicks away. You can read the details in the article, but the short version is that a professor and his graduate student at UCSD, in a moment of sheer inspiration, driven by deadlines, ripped open the centerfold of a then-recent Playboy and scanned a 5.12 inch portion of the page to get this widely circulated image.
After nearly 40 years since that brazen act – which was done without worrying about things like copyright and fair use – Deanna Needell and Rachel Ward decided that it was time to shake things up. “We need a new standard test image,” they reasoned. And the choice of a new image had to undo the wrongs of the past. You see, some in the community take offense to using Lena’s image given its source’s exploitative treatment of women. Some journals have gone to the extent of banning the image’s use in their papers! Culture wars comes to Image Processing!
So our friends set about correcting this historical wrong. They contacted the right people (modeling subject), obtained the right permissions (fair use) and used the new test image in their experiments. As good and responsible members of the research community, they are making the new image available to other researchers (along with fair use permission). What is the new image? Just to show that I’m not kidding, I’ll copy-paste from the pdf itself. So no kidding, and no complaints from me!
Here you go –
Credits: The sample set of results on Lena image is from this article. And the Fabio image is from Needell and Ward, “Stable image reconstruction using total energy minimization”.