Is mammalian vision processed as a sequence of frames?

Question

I often read that people believe that human vision has an inherent frames-per-second rate (FPS) that causes stroboscopic effects - such as seeing the spokes of a rotating wheel apparently rotating at a different speed or appearing stationary when moving.

For example: in a Physics.SE answer to Can a “superhuman” move so fast that an average person cannot see them?

The human eye-brain visual refresh rate has an "effective frame rate" of around 30fps.

Have you never watched the rims of the wheel in a car next to yours? I can clearly recall many times watching as the wheels sped up how it appears to stand still then move backwards.

This surprises me as I would expect there is no synchronisation of neuron firing in the retina, that neuron firing rates would vary widely depending on light levels and that the brain has no need to process the continuous signals in fixed cycles or in cycles whose length is invariant.

I believe this question, or a good answer to it, would differ from What is the equivalent of shutter-speed in Human eye?

Is human vision in any way subject to a fixed frame rate of 30 FPS?

Answer 1

Short answer: no, there is no fixed frame rate or frame-based processing in mammalian vision.

Photons arriving at the photoreceptors at the back of the human retina interact with photo-sensitive pigments called opsins, and modulate their release of the neurotransmitter glutamate . The level of glutamate released from a photoreceptor then changes the membrane potential of the other neurons in the retina connected to the photoreceptor (bipolar cells, for instance). Signals are processed within the retina, then passed through retinal ganglion cells (RGCs), along the optic nerve, through sub-cortical structures and in to visual cortex.

This process occurs continuously, but all of the elements along the way have intrinsic time constants and other temporal constraints (number of photons required for opsin activation; recovery rate of the opsins in the photoreceptors; membrane time constants and refractory periods of neurons; synaptic delays; etc.). Persistence of vision (the effect underlying the perception of a movie as a continuous stream rather than a series of distinct frames) probably relies on the conjunction of many of these time constants.

Concretely, a flashed visual stimulus produces a response in (primate) visual cortex after around 30–50 ms (Maunsell and Gibson 1992). You can recognise and subsequently recall a lot of detail from a sequence of images flashed at around 100–150 ms per image [citation needed]. These are still not hard and fast definitions of the "speed" of vision.

Since we only learnt to produce movies as sequences of distinct frames, it's tempting to guesstimate the speed of human vision in those terms. Biology, as always, is more complicated.

As far as the "spinning rims" issue goes, there are many aspects of cortical and subcortical activity that tend towards oscillatory activity. In a system with an intrinsic frequency (for example, the neurons in the thalamocortical loop between LGN and primary visual cortex), providing an oscillatory stimulus can synchronise the system to the stimulus.

Many of the intrinsic time-constants I described above could result in a tendency to oscillate at some intrinsic frequency, and then be subject to phase locking due to a periodic stimulus close to that frequency.

Answer 2

While in human made cameras light sensor and any image processor are separate devices, in the living eye the first steps of image analysis begin in retina already as light sensing cells are part of the nervous system. The optic nerve carries pre-processed information and not the encoded original image to the brain. This pre-processed information is not segmented into frames.

The "image processing system" of the human vision would generalize a presented sequence of frames into logical continuous movement, even if the frame rate is as low as few FPS only. This does not mean the eye cannot see that is going on. Frame rates as high as 120 FPS are required to reach the inertia limits so that the flickering really cannot be seen.