Graphics 2 Computational Theory of Colour Perception

1
Graphics 2Computational Theory of Colour
Perception
2
introduction

• summary - cognitive science and graphics
• evidence for a computational model of colour
  perception
• visual perception pathways
• receptive fields
• structure of the LGN
• opponent character of colour
• developmental psychology
• evolution of the eye
• the computational model
• discovery of colour systems and modern colour
  abstraction
• conclusion

3
cognitive science and graphics

• experience and training provide rules of thumb on
  how best to arrange graphic elements
• can the cognitive science, and the psychology of
  perception in particular, predict any of these
  rules?
• are there any low-level / computational
  approaches to the mind that can predict graphics
  principles?
• if so, could the study of cognitive science guide
  us towards better design, eg for signs and
  pictograms?
• example - mechanism of the perception of colour
• evidence from physiological and psychological
  studies of trichromatic perceptual system in the
  human eye

4
visual perception pathways
5
(No Transcript)
6
visual perception pathways

• optic tract delivers information to lateral
  geniculate nucleas and superior colliculus
• additional inputs from brainstem - attention,
  saccades
• input from visual cortex - feedback loop
• structure suggests LGN important in organising
  information
• visual cortex
• V1 area (or striate cortex) receives its input
  from LGN
• all higher perceptual processing occurs in the
  visual cortex - hence cortical blindness
  damage to the cortex results in blindness
• some evidence of pre-processing in the LGN -
  physiological structure, phenomena of
  blindsight, eg Weiskrantz (2007)

7
receptive fields

• experimental evidence shows physiological
  arrangement of groups of cells in the retina
• photoreceptor cells connect to bipolar cells
• bipolar cells connect to ganglion cells which
  form the optic nerve fibres
• outputs of collections of rod cells are combined
  before reaching the optic nerve

8
receptive fields

• ganglion receptive field organisation
• centre / surround antagonistic arrangement of
  cells
• excitatory / inhibitory nature of receptive
  fields
• receptive fields in fovea are very small -
  increase in size towards periphery
• small receptive fields allow higher resolution -
  greater visual acuity

9
receptive fields

• ganglion receptive fields overlap slightly
• a mixture of centre-off and centre-on cells send
  signals to the LGN
• precise locations in the LGN correspond to
  specific areas of the field of view
• eg DeVries and Baylor (1997) structure of
  ganglion cells in rabbit

10
(No Transcript)
11
Herman Grid - explanation
12
structure of the LGN
13
structure of the LGN

• cell layers 1 and 2 are known as magnocellular
  layers and are made up of nonopponent cells
• cell layers 3, 4, 5 and 6 are known as
  parvocellular layers and are made up of
  opponent cells (with a tiny fraction of
  nonopponent cells)

14
structure of the LGN

• nonopponent cells
• when light of any wavelength shines on retina,
  some cells increase and some decrease activity -
  nonopponent cells do not react to colour
• opponent cells
• increase or decrease activity depending on
  wavelength
• some increase with long wavelength and decrease
  with short wavelength (on / off response)
• some do the opposite
• light with both wavelengths has no effect
• one type switches on/off or off/on between red
  and green
• the other type switches between blue and yellow
• evidence from many studies, building on
  pioneering work of DeValois et al, (1960) Wiesel
  and Hubel (1966)

15
opponent character of colour

• colour complement
• we see reddy yellows, yellowy greens, greeny
  blues
• we seem never to see reddy greens, or yellowy
  blues
• colour contrast
• we seem to experience the same colour differently
  depending on its adjacent colours
• Leonardo da Vinci wrote about the effect in his
  Treatise on Painting (around 500 years ago)
• colour constancy
• we can distinguish colours in a very wide range
  of brightnesses
• we can usually perceive adjacent colours even if
  one is very much brighter than the other

16
(No Transcript)
17
(No Transcript)
18
(No Transcript)
19
opponent character of colour

• highly structured organisation of cells
• retinal cells -gt ganglion receptive fields
  on/off centre/surround
• ganglion receptive fields -gt LGN receptive
  fields on/off centre/surround
• LGN receptive fields -gt visual cortex receptive
  fields on/off centre/surround
• chromatic and achromatic systems
• evidence suggests processing in the LGN and
  visual cortex generates three channels
• Blue-Yellow channel
• Red-Green channel
• Achromatic channel

20
are colours inherent or learned?

• the three channel model requires that we
  experience colours inherently
• if colour perception is a result of underlying
  brain structure then there should be no need to
  learn to distinguish red, green or yellow, blue
• as babies grow and develop they learn the names
  for different colours
• how can we tell which came first?
• do we learn to distinguish colours as we learn
  language, or do we simply learn labels for
  something that we all experience almost from
  birth?

21
spectral response of cone cells
22
are colours inherent or learned?

• infant studies, eg Bornstein, (1976)
• 4-month old infants shown 480nm light (what
  adults would identify as blue light) until they
  lose interest
• then presented with either 450nm light or 510nm
  light
• 30nm shift to 450nm is still what adults call
  blue - subjects are still not interested
• 30nm shift to 510nm is what adults call green -
  subjects show interest
• suggests subjects perceptual experience has
  changed
• suggests no need for language in order to
  recognise the change of colour
• same effect demonstrated with other spectral
  ranges of light
• strong evidence that colour categories do not
  need to be learned
• hence 3 channel model supported by behavioural
  studies

23
evidence from evolutionary studies

• evidence from the evolution of vision
• older animals have achromatic vision only, eg
  marsupials
• more recently evolved animals have dichromatic
  vision, eg squirrels
• most recent have trichromatic, eg apes
• evidence from the field of phylogeny
• colour vision systems have evolved independently
  several times during evolutionary history (fruit
  flies, sticklebacks, gorillas)
• when colour perception evolves, it seems to
  follow the same sequence
• achromatic - light from dark
• dichromatic - achromatic plus blue from yellow
• trichromatic - achromatic plus dichromatic plus
  red

24
computational model of colour perception

• evidence from the structure of the retina
• receptive fields of ganglion cells
• on/off centre/surround functionality - Hermann
  grid
• opponent character of colour perception
• colour complement, colour contrast
• colour constancy
• evidence from developmental psychology
• evidence from the evolution of vision
• progression from achromatic to trichromatic
• functional model
• represents the overall computational mechanism
  from retina to visual cortex supported by
  evidence above

25
computational model of colour perception
26
Newtons experiments with colour

• A new theory of light and colours (1672)
• repeated earlier prism experiments of Marci
  (1648)
• discovered that the spectrum could be recombined
  into white light
• concluded that there were seven distinct pure
  colours and that all other colours were obtained
  by mixing these pure colours in different amounts
• produced his circle of colour that did away with
  black and white and the old linear representation

27
Goethes Theory of Colours

• More than 2000 pages of notes published 1808-1823
• Colour Circle developed 1793
• disagreed with Newton (however he was approaching
  the field from a psychological viewpoint)
• based his conclusions on empirical study of
  people
• concluded that there were three primary
  colours, and three complementary colours

28
modern approach to colour abstractions
29
modern colour abstraction systems

• contemporary design usually 12 colours
• primary
• secondary
• tertiary
• others can be used
• various print production systems
• Pantone
• 4 colour process
• Hexachrome

30
conclusion - predictions

• computational theory of opponent colour predicts
  four colour categories
• Reds, Greens, Yellows, Blues
• predicts contrasting and complementary colours
• We see reddy yellows, reddy blues, greeny
  yellows, greeny blues
• We seem to never see reddy greens, or yellowy
  blues

31
conclusion - explanations

• colour constancy explained by separate achromatic
  channel
• visual cortex is able to process colour and
  brightness information separately before final
  perception occurs
• hence comparisons can be made with other areas of
  the field of view
• saccades may enable data from the field of view
  that lies outside the fovea to be used to
  calibrate brightness

32
conclusion - explanations

• we find bright red adjacent to bright blue and
  bright yellow adjacent to bright green unpleasant
• all three channels highly active - feedback loop
  invoked?
• we find shades and complements pleasing and
  interesting
• moderate activity throughout the pathways of
  visual perception?
• other aspects of colour theory
• inherited colour blindness?
• missing or abnormal L-type 2 of males, 0.04
  females
• missing or abnormal M-type 6 of males, 0.4
  females
• missing or abnormal S-type very rare
• ?

33
colour wheels and colour pickers

• http//wellstyled.com/tools/colorscheme2/index-en.
  html
• generates colour schemes from a base colour and
  contrast/complementary colours
• http//r0k.us/graphics/SIHwheel.html
• interactive colour wheel, plus lots of
  information about colours and colour theory, and
  links to other resources
• http//www.colorschemer.com/ColorPix.exe
• handy tool for identifying colours on screen

34
further information

• Virtual Colour Museum
• http//www.colorsystem.com/
• Webvision
• http//webvision.med.utah.edu/
• Webexhibits
• http//webexhibits.org/colorart/

35
references

• MH Bornstein, W Kessen, and S Weiskopf (1976)
  The categories of hue in infancy Science, Vol
  191, Issue 4223, 201-202
• Steven H. Devries and Denis A. Baylor (1997)
  Mosaic Arrangement of Ganglion Cell Receptive
  Fields in Rabbit Retina The Journal of
  Neurophysiology, Vol. 78 No. 4 October 1997, pp.
  2048-2060
• Russel, DeValois (1960) Color Vision Mechanisms
  in the Monkey J. Gen. Physiol., Vol. 43, Issue
  6, 115-128, July 1, 1960 accessed 10 Feb 2008
  from http//www.jgp.org/cgi/content/citation/43/6
  /115
• Larry Weiskrantz (2007) Blindsight
  Scholarpedia, 2(4)3047. http//www.scholarpedia.
  org/article/Blindsight
• T. N. Wiesel and D. H. Hubel (1966) Spatial and
  chromatic interactions in the lateral geniculate
  body of the rhesus monkey The Journal of
  Neurophysiology, Vol. 29, Issue 6, 1115-1156,
  November 1, 1966 accessed 10 Feb 2008 from
  http//jn.physiology.org/cgi/content/citation/29/6
  /1115
• Sekular and Blake (1990) Perception McGraw Hill
• Lythgoe, J. N. (1979) The Ecology of Vision,
  Oxford University Press
• Gordon, I. E. (1989) Theories of Visual
  Perception Wiley
• Humphrey, Nicholas (1992) A History of the Mind,
  Chatto Windus.