Been doing some testing to find the flow of pixels in portraits. Above are a few attempts to find vectors in pixels based on surrounding pixels. In video this is achieved by simple comparing current frames to previous frames then calculating the velocity of the different pixels. Here I want to find the general flow of pixels through the image. The goal is to then have particles stream through the image to generate a representation of the image, some type of generative stipple drawing.
See large image here
Original Einstein image here
Original Obama image here
share it
tweet it
February 3rd, 2010 at 10:16 am
Hi!
Very nice, would be real fun in a live installation :)
Could you a say a bit more in how you’re computing the “pixel flow”?
Is it like optical flow?
Any refs would be appreciated.
February 3rd, 2010 at 12:20 pm
That’s insane. It’s beautiful!
February 3rd, 2010 at 12:53 pm
OK — I am not an artist, so I won’t use the right wording here, but…
In art and architecture, one often speaks about how the eye moves over a piece. Techniques like chiaroscuro are used to focus the viewers’ attention, etc.
So the question is this: How well can you use your pixel flow to predict visual flow? Or, alternately, how might a record of eye tracking on an image allow you to back-calculate the pixel flow?
February 3rd, 2010 at 12:58 pm
Dan, That is really interesting. I would love to do an eye tracking experiment on a work of art and then compare that to the pixel flow. Lets get on this :)