Is human vision performed concurrently or step-by-step

[sazr]

Is most of human vision processing performed in a linear step-by-step fashion in the brain or is it concurrent? Let's take for example; for the process of us recognising a pen on a table.

• Light/information enters the eye, is focused on the fovea part of the retina because we are focusing on the pen on the table. This also means that many cones are being excited and emitting a signal (plus some rods are also receiving light and emitting a response I imagine?).
• The eyes/rods/cones do some sort of compression or pooling before emitting their signals via the ganglia through the optic nerve to the visual cortex. I imagine both rods and cones do this because there are 126 million photosensitive cells and only 1 million ganglia.
• The primary visual cortex starts finding edges and other low level stuff. Does this get performed before the secondary visual cortex or concurrently?
• The secondary visual cortex begins processing for object size, colours, shapes. The ventral system is involved here. Is the dorsal system involved aswell? Does the ventral system require the processed edge information (from the primary cortex) in order to do its work? Is its processing concurrent?
• Finally the ventral system recognises that there is a pen on the table.

Are you able to correct any mistakes I have above in the visual processing pipeline? Can you point out what steps in this pipeline are performed concurrently (if any) and what are step-by-step (and what that order of operations would be)?

Any advice about this process would be extremely helpful :)

 

[BillTre]

The retina is a large sheet of cells that passes signals in parallel (after a few processing steps to the retinal ganglion cells that go from the retina to the brain. There are several layers of cells in the retina that do some signal processing before it ever gets to the output cells, the retina ganglion cells. There are things like center surround detection and line detection in the retina, as well as probably adaptive functions for things like lighting level.

The ganglion cells project to the mid brain. Different sets of ganglion cells can project to different places in different species. In this way different signals can go to different areas of the brain. This is a pattern that can be repeated at different sets along the way. Different signals can end up going to several different places in the cortex, so this is kind of parallel but doing different things. However, each step is often a set of several cells, often retinotopic preserving the relationships of the visual field. This would be very parallel.

Any function like identifying an more complex object like a pen in a visual scene might involve building up images of the object and field through successive scans of the visual area (which often follow the lines in the scene). This would take longer and involve observations repeated in a short period of time to generate an internal "image" of what the visual system detects. Successive inputs would then update the scene and the objects.

 

[atyy]

Light/information enters the eye, is focused on the fovea part of the retina because we are focusing on the pen on the table. This also means that many cones are being excited and emitting a signal (plus some rods are also receiving light and emitting a response I imagine?).

Yes

The eyes/rods/cones do some sort of compression or pooling before emitting their signals via the ganglia through the optic nerve to the visual cortex. I imagine both rods and cones do this because there are 126 million photosensitive cells and only 1 million ganglia.

Yes, but the convergence is different for rods and cones: https://www.ncbi.nlm.nih.gov/books/NBK10850/.

The primary visual cortex starts finding edges and other low level stuff. Does this get performed before the secondary visual cortex or concurrently?

It is ok to think that the primary visual cortex does it work first, then passes the processed information to higher visual cortical areas. It is conceptually not much different from the processing done in artificial neural networks with many layers, also referred to as deep learning.

The secondary visual cortex begins processing for object size, colours, shapes. The ventral system is involved here. Is the dorsal system involved as well? Does the ventral system require the processed edge information (from the primary cortex) in order to do its work? Is its processing concurrent?

It is mainly the ventral stream, though I'm not sure what would happen if you removed lots of the dorsal stream.

Finally the ventral system recognises that there is a pen on the table.

I'm not sure where the "finally" or the little man in your head is.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3306444/