>Short answer: Reproducibility. The image is one of several which have been commonly used in the literature for decades.
If we are going by that standard we would probably have to use something like BBC Test Card F
>Also, it's actually, it's a really good photo for testing various types of computer vision algorithms - complex backgrounds (including a mirror), varying textures and colors (e.g. the hat feather thing), and a simple grayscale conversion works well.
Not really. Firstly we don't have great documentation on the hardware involved. The hardware involved is also highly non typical by modern standards (if your system is designed around drum scanned film and its 2015 you are doing something wrong). Using an image produced by a digital camera using Bayer filter would be more fit for purpose (the most obvious difference is that you will get more information on the green channel that the red and blue channel, this doesn't happen with film) . It would also avoid the rather messy copyright situation around the image.
Its not 1973 we are not short of digital images to use for testing any more. Chosing one that doesn't come from playboy is a reasonable choice.