8

In electronics the Nyquist sampling rate shows you the maximum frequency that could be sampled without aliasing. So if the human eye is able to see 24fps does it mean that you can see all the information below 12fps?

suyol854
  • 310
  • 1
  • 4
  • 13
  • 1
    Are the brains digital? Do we have a clock inside, so all operations are synchronous like in the computer? Which numerical base have our brains? Or maybe they are not digital at all, they could be analog computers, series of comparators and amplifiers. Who knows? But one is certain, the brain is far more complex than a computer. – Marko Buršič Feb 11 '18 at 21:36
  • 15
    The hypothesis (that the "time resolution" of human vision is 24 fps) is false anyway. – R.. GitHub STOP HELPING ICE Feb 11 '18 at 22:05
  • 1
    @MarkoBuršič Actually, the brain does have a sort of clock to synchronize various neural activities. This is what brain waves are. Unlike a computer clock signal, the waves do not perfectly synchronize each neural, but rather allow various regions of the brain to process information at the same time. – forest Feb 12 '18 at 03:15
  • 2
    I can see and be annoyed by display flicker at frequencies much higher than 24 fps. – user57037 Feb 12 '18 at 06:08
  • 1
    I'm voting to close this question as off-topic because the human eye is not electronic. – pipe Feb 12 '18 at 10:31
  • 5
    Really interesting question but not here, voting to close as off topic. – RoyC Feb 12 '18 at 12:15
  • The human eye has a low pass filter. – user253751 Mar 05 '18 at 04:34
  • 1
    Quite a bit of psycho-visual analysis here, but remember "24 fps" implies that the information was sampled at that rate before it ever reached your eyes. – gbarry Nov 11 '18 at 06:56

4 Answers4

18

The firing of neurons in the eye is completely asynchronous, so for all intents and purposes, the process we call "seeing" must be considered continuous-time, not sampled. Nyquist does not apply.

If things move too fast, they simply appear blurry. And "too fast" varies from person to person; there is no well-defined cutoff frequency for the general population.

Dave Tweed
  • 172,781
  • 17
  • 234
  • 402
  • 7
    I agree with the "sampling rate is a concept that doesn't apply to the eye" statement. We can, however, even in continuous time, make information-theoretical statements: If e.g. a neuron has a very nonlinear transmission function (ie. there's both a max as well as a min number of firings per second that still "mean a thing"), then that simply distorts the transinformation between source (image) and sink (brain) and we can make statements like "An image that changes x times per second contains more b bits of information, which is much more than the transinfo capacity of the observational system – Marcus Müller Feb 11 '18 at 19:37
  • 5
    …, and thus, information loss must occur". Mind you, coming up with numbers here is near impossible. – Marcus Müller Feb 11 '18 at 19:38
  • 8
    And to make things worse, a single eye-ball has various regions of greater or lesser sensitivity to changes in intensity. Personally I can detect a distinct flicker in a 60 Hz (CRT) computer screen when viewed out of the corner of my eye, that I cannot detect when I look directly at it. –  Feb 11 '18 at 19:45
  • 2
    @Wossname: Yes, that's the difference between the fast "rod cells", which are sensitive to intensity only and distributed more toward the periphery of the retina, and the slower "cone cells", which are sensitive to color and concentrated more in the center (fovea). I happen to be working with some retinal implant researchers at the moment, and they consider the ability to stimulate the neurons with a time resolution of 1 ms "barely adequate". – Dave Tweed Feb 12 '18 at 02:40
  • 1
    The human brain visual-cortex has the task of initializing and reliably updating a 3_D model of the world around us. As we move, the optic-flow of interesting features being occluded and later revealed is continually validating or reducing confidence in the 3_D model. If we simply sit and sway back and forth, the binocular and monocular disparity similarly reinforce or challenge the 3_D model. All this reinforce/challenge behavior requires information. – analogsystemsrf Feb 12 '18 at 06:00
  • The statement about neurons being "completely asynchronous" is false. Neurons are not completely synchronous, but they are subject to periodic phenomena such as brain waves. – Dmitry Grigoryev Feb 12 '18 at 08:10
  • Aliasing can still occur for non-uniformly sampled data. Since the neurons sample (“fire”) at some time sequence, the process should be treated as digital. – user110971 Feb 12 '18 at 08:47
  • The eye doesn't sample in a "frequency". Each of its cells are at any point in time activated to some degree. This idea of 25 / 30 frames per second thing is because change in activation doesn't happen instantaneously as the light input to a cell changes, but rather takes about a 30th of a second (with lots of different factors affecting exactly how long). So if any change happens faster than that, the cells in our eyes just can't change their activation levels quickly enough. – Arthur Feb 12 '18 at 09:09
  • @DmitryGrigoryev: We're talking about the eye, not the brain. Besides which, brain waves are an emergent phenomenon that only appears when you measure the collective activity of very large numbers of neurons. It doesn't imply anything about the firing behavior of any particular neuron. – Dave Tweed Feb 12 '18 at 12:48
  • @user110971: Digital (quantized in terms of voltage or some other quantity) does not necessarily imply that time is also quantized. There are plenty of continuous-time digital systems. – Dave Tweed Feb 12 '18 at 12:52
  • @DaveTweed Actually many brain waves are not emergent but have a specific locus where they are generated. An example is the temporal binding produced by recurrent thalamocortical resonance activity. This helps synchronize the thalamocortical system. – forest Mar 05 '18 at 04:09
  • @user110971 It's the rate of firing that transfers information (technically, firing rate is just one way that neurons encode information). The rate itself is what matters, not whether or not an action potential is active at a given time period (which would make it digital serial communication). It is similar to an analog circuit in that respect. The reason people often conflate it with digital communication is because an action potential is an all-or-nothing event, so it's easy to incorrectly conclude that firing = 1, no firing = 0. – forest May 10 '18 at 03:51
6

The terminology used when applying to the human eye is the Critical Fusion Frequency, also called the Flicker Fusion Threshold. This is the frequency at which a point of flashing light merges visually into a single solid point of light. The precise frequency differs depending on what area of the retina is being stimulated, how bright the light is, how much ambient light there is, and what wavelength the light is. At the lowest (dim, blue light, center of retina), the frequency can be as low as 10 Hz. At the highest, it can be over 60 Hz. A saccade (rapid eye movement) temporarily disables this limit.

This is not a limitation of the retina itself, but rather an intentional blurring that occurs in the brain called visual persistence, a phenomenon that allows us to process briefly-seen images for a split second longer, at the expense of lower temporal resolution. The reason subliminal visual stimuli work is because an image can be flashed slow enough that our retina can pick it up, but too fast for our visual persistence to kick in, leading us to have no conscious awareness of the image having been displayed. An image of a couple kissing that flashes for a matter of milliseconds will elicit positive emotions, despite the fact that our visual persistence does not give our conscious mind enough time to process the image and be aware of it.

For the human visual system, an excessive frequency will not result in aliasing. Aliasing applies only to a periodically sampled signal when the original signal contains excessively high frequencies. The visual system does not have a discrete sample rate.

forest
  • 518
  • 3
  • 15
  • 1
    Re, "aliasing is applied specifically to digital images." Actually, aliasing happens any time you are dealing with a periodically sampled signal, and the original signal contains frequencies that are higher than the Nyquist frequency. The signal does not necessarily have to be an image, and the "frequency" does not necessarily have to be a frequency in the time domain. – Solomon Slow Feb 12 '18 at 11:17
3

There is a known visual illusion where objects moving periodically at a certain frequency appear to be moving in the opposite direction under continuous illumination. There are different theories explaining such observations, and one of them suggests that human vision has what could loosely be called "sampling frequency", somewhere between 15 and 20 Hz. AFAIK there is still a debate on the matter, and since classic experiments on subjective stroboscopy involved taking LSD and similar drugs, experimental data in this area is far from complete.

What you will certainly not find is a sample frequency which is constant across population, different light conditions or even different parts of the visual field.

Dmitry Grigoryev
  • 25,816
  • 5
  • 45
  • 107
0

Sampling can be simulated by blinking. Imagine a clock with one arm that you can set to rotate at arbitrary speed in either direction. If the arm rotates more than 180 degree when your eyelid was closed, you cannot say which way it rotated. This is exactly what aliasing is.

rakeshdn
  • 9
  • 2
  • Obviously the user already knows aliasing is, hence the question. This answer only explains what happens if you introduce artificial sampling, which isn't the question. – pipe Feb 12 '18 at 15:31