In mid-2013 I became fascinated by pixels and colors in digitized images. Specifically, I wondered if the proportions of colors present in digitized images had any bearing on our experience of the paintings. Could different proportions of colored pixels have as much effect on a viewer as the form or content represented by the pixels? In my first experiment, i sought to rearrange all the pixels of an image in such a way that would show the proportions of colors without the form or the content of the image. I quickly discovered that color is represented three-dimensionally in the digital world while an image is a two-dimensional. As result, there is no way to organize pixels by color in a 2D space without having breaks and artifacts. So I decided to decompose a digitized version of Kandinsky's Composition VII by hue, saturation, and brightness.
After Kandinsky, I began decomposing every image I could find and studying the proportions of colors while looking for meaning. Needless to say, I found none. Then one night I was watching "The Last Airbender" with my girlfriend. "This movie sucks!", I exclaimed, "But because some of the scenes have such beautiful colours, I want to keep watching". "We can watch it on mute", she suggested. "I wish I could take the colour proportions from some of the scenes in this movie and put them on other footage", I said aloud. Then I had flash and ran to my computer. "What if I rearranged the pixels of an image based on the color proportions of another image???" And then I did exactly that. Suddenly, an experiment had become an artform. Pixels shifted like blowing sand to reveal new images. I specifically chose a digitized Picasso painting because of its color diversity and a car advertisement because of the smooth changes between colors.
After the Picasso to Picasso experiment I thought I could prove that given a certain resolution, the pixels of any digital image could be rearranged to form a reasonable representation of any other digital image. It seems to hold true for all digital images of photographic origin, but vector graphics simply don't have enough color diversity to represent many other images. So sticking with images of photographic origin I began to search for ways to better rearrange the 3D color data of pixels in the 2D space of the image. I learned about Hilbert Curves and 3D LUTs but eventually settled on my own low-fi methods. Rather than sorting pixels two-dimensionally, I would sort them first one-dimensionally(treating the entire image as one long row of pixels) and then I would sort each columns individually. The results were considerable better than my previous experiments so I decided I was ready to go from images to video. Unfortunately, the new processes were brutally slow and one minute of footage took about 30 minutes to render. BUT! I could finally watch and compare color decompositions of video!
The next step was clear. I made a two part video. First, rearranging the pixels of each frame of a video to form the image of of a fractal, then rearranging the pixels of the image of a fractal to form each frame of a video. I was astounded how the form and content of the resultant image/video was completely unchanged! It looked as if someone had just applied some image processing effect when in reality every pixel was being rearranged every frame.
So where to now? Maybe some hi-res prints of sorted pixels of famous images in a gallery? I'd like to see the blowing sand effect at a 4k resolution and of course I still need to rearrange the pixels of every frame of a video to form the frame of another video. I also think this technique has a lot of potential for a music video. Contact me ;)
