The Histogram as the Image:

Even better: self-similar histograms!

Tags: , ,

9 Responses:

  1. taffer says:

    That's the most impressive thing I've seen all day.

    In other news, I don't get out much.

  2. anyfoo says:

    Wow. Best idea in a very long time. I love it.

  3. jesus_x says:

    I saw the first one a few days ago on another blog, but the other two rock.

  4. ralesk says:

    Epic awesome.

  5. thedimka says:

    it is interesting to do with matching images.

    I actually thought in the past about embedding secret messages into histogram, but skyline is more fun

    • ... but not very useful. Steganography like this is cute, but not exciting. 4 bits at a time in arbitrary (but known) locations per kilobyte (meaning two KB of image for one byte of data) in a grainy, blown-up picture of Paris Hilton getting out of a limo that, when reconstructed, make a few sentences of 7-bit ASCII are useful and exciting.

  6. siobhan1 says:

    So darn neat!

  7. telecart says:

    The only problem is that besides skylines and maybe a few kinds of landscapes, there's not much art in histograms.

    Still, pretty nifty for what it's worth.

  8. greatbiggary says:

    Nice. When I saw the IronicSans post last week, I wanted to do exactly what Josh Millard did, but I've been way too busy to sit down with it.

    Back in grade school, bored at my stepdad's office, playing with the photocopier again, I had an idea to create an image that used different colors that were all the same gray value when photocopied - an image that couldn't be copied by the technology of the day. I didn't have the patience to play with paints, or whatever else I'd need.

    A few years later, playing in Qbasic, I revisited the idea, poorly, and got nowhere. Last year, I remembered the idea. In Processing, I made a 4096^2 image of every 24-bit color. Side note: scrambling the pixels randomly resulted in an apparently evenly middle-grey square when viewed at a distance. Then I brought the image into The GIMP, converted to grayscale, and had Processing compare same-brightness pixels with the original image to extract all original colors of the same luminance to new images.

    This is all value 63

    All 127:

    All 191:

    The next step would be to map these shimmering values to color images, which would hilariously seem contentless to the fully colorblind. Of course, the RGB->gray conversion may be different in other packages, and different eyes/sensors will pick things up differently, so it can perhaps never be perfect, but it's interesting.