Tuesday, April 02, 2019

The physics of vision

We had another great colloquium last week, this one by Stephanie Palmer of the University of Chicago.  One aspect of her research looks at the way visual information is processed.  In particular, not all of the visual information that hits your retina is actually passed along to your brain.  In that sense, your retina is doing a kind of image compression. 

Your retina and brain are actually anticipating, effectively extrapolating the predictable parts of motion.  This makes sense - it takes around 50 ms for the neurons in your retina to spike in response to a visual stimulus like a flash of light.  That kind of delay would make it nearly impossible to do things like catch a hard-thrown ball or return a tennis serve.  You are able to do these things because your brain is telling you ahead of time where some predictably moving object should be.  A great demonstration of this is here.  It looks like the flashing radial lines are lagging behind the rotating "second hand", but they're not.  Instead, your brain is telling you predictive information about where the second hand should be.

People are able to do instrumented measurements of retinal tissue, looking at the firing of individual neurons in response to computer-directed visual stimuli.  Your retina has evolved both to do the anticipation, and to do a very efficient job of passing along the predictable part of visualized motion while not bothering to pass along much noise that might be on top of this.  Here is a paper that talks about how one can demonstrate this quantitatively, and here (sorry - can't find a non-pay version) is an analysis about how optimized the compression is at tossing noise and keeping predictive power.  Very cool stuff.

3 comments:

  1. This is one of my favorite topics in physics. In particular, when I was a graduate student, arguably the most memorable paper out of the thousands I browsed and read over five years was this one: http://threeplusone.com/Still2012.pdf.

    ReplyDelete
  2. PPP, thanks! That paper is very interesting - I wish I had time right now to read it carefully. It's now on my list.

    ReplyDelete
  3. It gets even weirder... There is post-diction in addition to prediction. Take a look at this work by David Eagleman:

    https://www.eagleman.com/papers/eagleman_sejnowski_2007.pdf

    ReplyDelete