Naked Science Forum
General Science => General Science => Topic started by: sazr on 16/05/2019 01:23:20
-
A while back I asked about human vision processing and if it is performed synchronously or asynchronously. One of the responses mentioned that the retina performs 'Center Surround Detection'. So I tried to google what this is but no relevant results came up.
Is anyone able to explain what this is and how its useful or point me to a link that explains it (that is not too high level/hard to understand)?
-
Perhaps they were referring to a technique using operations on a 3x3 set of pixels which compare the levels at the center and surrounding pixels to detect edges in the image?
- I understand that applying the same matrix to all pixels of an image is called a convolutional neural network (https://en.wikipedia.org/wiki/Convolutional_neural_network)(?)
Of course, neurons in the retina can reach a lot wider than just 3x3 "pixels" (or rods and cones, in the case of the retina).
- And rather than apply the same matrix operation to all points on the image, the thing most important is to detect the things that are expanding in your field of vision: The tree that you are about to hit, or the lion that is about to eat you..
- This means that you would apply a different matrix depending on where it is in your field of view
- and neurons can do time-based analysis in order to detect moving edges, which a 3x3 pixel filter cannot.
Anyway, there are multiple methods for doing this:
See: https://en.wikipedia.org/wiki/Edge_detection
-
Also conceivable that it refers to the higher spatial and color resolution in the center field of the eye. Peripheral vision is good at detecting movement but doesn't discriminate color or detail well, so we may be alerted to "something" offcenter and swivel quickly towards it to build up a detailed image.