3 - PROBLEM WITH COLOUR VISION/GREY WORLD

Question 1

problem with colour vision in terms of spectral reflectance

Answer 1

‘getting information about spectral reflectances out of retinal images, or more precisely, out of the spectral (wavelength) content of retinal images’

Question 2

formula and problem with it

Answer 2

S = spectral content of surface’s image (proximal stimulation - reflected light)

R = spectral reflectance of the surface (distal stimulus)

I = spectrum of illumination

S = R x I

R = S/I

^ can’t do this second equation as we don’t know I (illumination spectrum)

• this is the problem with colour vision
• ‘but visual system must be solving this for us because we have colour constancy’
• ‘the solution is not quite perfect because our colour perception is not quite constant in all illuminants, but it’s pretty good’

Question 3

how does the visual system overcome this problem?

Answer 3

no complete answer to this

but one idea = grey world theory

• but it is known that this is not the exact way that the visual system overcomes the problem
• the visual system must ‘somehow discount the contribution the spectrum of the illumination is making’
• ^ ‘dividing out the illuminant spectrum from the reflected light spectrum’
• but need to know what it is in order to divide it out
• so the visual system must estimate the illuminant spectrum so it can then divide out the reflected light spectrum
• it somehow manages to estimate the illuminant spectrum from the light reflected off numerous surfaces
• supposedly through the grey world theory

Question 4

grey world problem - what it highlights

Answer 4

‘perceived colour of a surface depends not only upon the light in the image of that surface, but also upon the light in the images of all the other surfaces contributing to the retinal images’

basically the colour of an object is determined by the colour of the whole image and not just the object of focus

Question 5

grey world assumption

Answer 5

‘if you average all the reflectances in a typical scene, then the resulting average spectral reflectance curve is close to flat (achromatic/grey)

• could average across the whole retinal image
• ‘colour could then be computed by dividing the signals from the image of a small patch of retina by this average signal (for an individual object)
• but would also predict failures in colour constancies in environments that are abundant in one colour (eg a very green forest)
• ‘grey world theory predicts that colour constancy will fail if scenes are not grey on average’
• ‘if the illumination changes, then the colour should change in predictable ways in non-grey scenes’
• ‘the predicted changes are not observed in human colour perception’
• would also assume that colour would be better perceived in natural illumination and would be perceived differently in artificial/fluorescent lighting
• also that ‘colour constancy breaks down under narrow band illuminations ie long wave red light only’
• and fluorescent lighting
• either look the colour of light or achromatic
• failed colour constancy

HUMANS DONT USE GREY WORLD THEORY FOR COLOUR CONSTANCY
- but it is a simple illustration

Question 6

‘if we look at a single patch of surface in isolation (surroundings = darkness) it’s colour will change as its illumination changes’

‘but it won’t if it’s part of a visible scene’

‘thus the brain uses light in many different regions of the retinal image in order to assign colour to any particular region’

‘proven by the fact we are NOT ‘blue blind’ in central vision, despite having no S-cones

Answer 6

-