Enlightenment bleg

For a research project, I’m trying to code up the lightness/darkness of about 100 black-and-white headshots of people. All the photos are currently in PDF, but could be exported into any other format.

Does anyone know of software that would – say – code a white box as 0, a black box as 100, a box containing a picture of a lighter-faced person at 30, and a box containing a picture of a darker-faced person at 50? None of the regular Adobe programs seem to have such a feature.

I could do this by hand, but it would be much less precise.

This entry was posted in Web/Tech. Bookmark the permalink.

6 Responses to Enlightenment bleg

  1. One alternative might be to try something similar to the approach used by Dan Hamermesh in at least some of his beauty papers. Hire some research assistants to rate the photos and use their ratings to construct an aggregate index. Whether or not this is a suitable approach will depend on the purpose of the ratings and the cost of hiring research assistants to do this among other things.

  2. SR says:

    Mathematica can do this fairly easily.

    Once the code has been setup, you can probably get 100 images done in a few minutes. The code will probably be fairly simple, only 10 lines or so
    (since all the code is preloaded).

    I’ve had a quick look the site, have a look at this to see some of the functionality: http://www.wolfram.com/products/applications/digitalimage/quicktour/pointoper.html
    This is the closest existing code available: http://library.wolfram.com/infocenter/Demos/399/

    You could probably modify it easily to get an average intensity and then scale to 100.

    Hope this helps,

  3. Andrew Leigh says:

    SR, thanks – though this may be beyond my technical/financial abilities.

    DE, funny you should mention that. Watch this space.

  4. Bruce Bradbury says:

    For only 100 (or even 1000), I would code by hand. The brain is much better at ascertaining skin tone than a computer.

    You need to take account of the exposure of each picture, the contrast, the gamma (roughly, contrast in the mid-tones) and maybe other factors such as the lighting angle (shadows). If you took the photos in a common studio with fixed exposure settings it might work.

  5. How uniform are the conditions under which the pictures were taken, and do you just want an overall assessment of the level of brightness in the picture, or are you specifically trying to identify the darkness of the individual’s skin?

  6. Andrew Whitby says:

    I agree with others that the human approach is probably better. If you have a lot of photos, it could be a job for the Mechanical Turk:


Comments are closed.