6
$\begingroup$

Coming from a computing science background, I noticed that cameras have orders of magnitude fewer wires than pixels. For example, the Raspberry Pi Camera v2 has 8 megapixels, but only 10 wires connecting it to the board. In human-made systems, it is preferred to time-multiplex pixels over few wires or, even better, already compress or encode image information on the camera.

I was wondering if the human optical system features something similar. Does the eye compress, encode or time-multiplex the image "pixels" it captures before sending the information to the brain? Or is each cone cell ("pixel") connected directly to the brain?

(In case the latter: Wow! The optical nerve must have a huge bandwidth.)

$\endgroup$
  • 2
    $\begingroup$ This isn’t my field so hopefully someone can give you a better answer, I just wanted to say that there is some processing in the retina. The signals from many photoreceptors are integrated by and/or into retinal ganglion cells, whose axons make up the optic nerve and number much less than the total number of photoreceptors. $\endgroup$ – canadianer Nov 29 '17 at 3:02
5
$\begingroup$
Are all cone cells connected directly to the brain?

No. Cones connect to ganglial cells.

Let's explore the numbers. Our eyes contain about 6 million cone cells, 90 million rods, and comparative handfuls of additional photosensitive cells (ipRGC's) that don't play a role in vision.

The optic nerve is a bundle of roughly 1 million fibers (perhaps up to about 1.7 million). So by just running the numbers, it doesn't fit.

I was wondering if the human optical system features something similar. Does the eye compress, encode or time-multiplex the image "pixels" it captures before sending the information to the brain?

Let's focus on photopic vision (color vision, which happens when lighting conditions are sufficient).

Our perception of color starts not with the cones, but with ganglial cells in the retina. This is different than how an RGB camera works; our brains don't process color in terms of "pure cone signals". Instead, color is processed in three channels; a "red-green" channel that measures the difference between L cones and M cones; a "yellow-blue" channel that adds input of L and M cones and subtracts out contributions from S, and a "brightness" channel that computes the sum of all cones. This is referred to as opponent processes. Maybe we can call this encoding.

In addition to this, there are ganglial cells in our retina that contribute to edge detection. These cells work similarly to ganglial cells that perform difference measurements for opponent processes, only they connect to the same cone types. Since ganglial cells connect locally, if there's a large difference between stimulation of two cones next to each other, it's an indication that there's an image of an edge of something projected onto that spot. If we can't distinguish individual cones, but these things can distinguish edges which to us are the more interesting property, maybe we can call this lossy compression.

Outside of this, a huge amount of our brain is devoted to visual perception. There are some really strange quirks about our eyes; most popular is the existence of the blind spot. Additionally, our 90 million rods are squeezed more into peripheral vision, leading to fuzzier imaging in photopic conditions the farther away we are from the fovea. Furthermore, chromatic aberration can be quite a pain, but L and M cone sensitivities are very close to each other. So where absolute sharpness matters, we have less S cones around; in fact, in our foveola, there's nothing but L and M cones. Our brain makes up for this, by stimulating us to move our eyes constantly and rapidly (saccades). As we do so, the visual cortex goes into overdrive, filling in that blind spot, crisping up that periphery a bit, painting in an illusory blue in our center vision, etc. Though this involves "moving the camera around", we could maybe call this a sophisticated, high level form of time multiplexing.

Much of this information is available from wikipedia articles, though you might need to know what to look for; so here's a starting guide:

$\endgroup$
  • $\begingroup$ Fascinating to read your answer! We basically have a Sobel operator in our retina. $\endgroup$ – user1202136 Dec 4 '17 at 8:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.