Colour constancy is often described as a computational problem the human visual system must solve. What is colour constancy and how is its solution achieved in the nervous system?
Colour constancy occurs when an object is illuminated by a coloured light and our visual perception continues to detect the objects apparent colour (Smithson, 2005); this is a postreceptoral process. The abstract colour coding can be mapped on a spectrum from black to white of which the receptive fields in the retina are registered to detect. Yet regardless of hue, brightness or saturation an illuminant may be shining on an object our visual system can compensate for the coloured light enabling the visual cortex to discern the colour of an object (Smithson, 2005). This has been described as a computational problem the human visual system must solve. Not all illuminants can be compensated for and the composition of light must contain a range of wavelengths our visual system can understand to discern an objects true colour (Smithson, 2005). Surface reflections of objects that are illuminated also play a role in how constant a colour appears. The basic area of the brain involved in the visual system begins with the retina which contains photoreceptors (cones (red (L), green (M) and blue (S) and rods) that enable us to see colours in dark and bright settings (Brainard, 2009). This is connected via the receptive ganglion cell (optic nerve) that relays visual stimuli to the Striate Cortex in the Occipital Lobe which is then directed to the Lateral Geniculate Nucleus in the Thalamus (Brainard, 2009). The following paper will discuss what colour constancy is and how the visual system achieves the ability of colour constancy.
There are two functions the visual system must process when the retinal field receives stimuli: an illuminant and object surface (Brainard, 2009). Illuminants are described as wavelengths that are controlled by a spectral power distribution for example day light which is ever changing in colour and distribution (Brainard, 2009). Surface reflectances are also depicted in wavelengths which are defined by the physical properties of an object (Brainard, 2009). Therefore surface reflectant wavelengths combine with illuminant wavelengths that stimulate the receptive field in the retina which involves postreceptoral processing to enable colour constancy (Arnold, 2009). L cones are red and stand for long wavelength, M are green and refer to medium wavelength, S are blue and refer to short wavelength. Wavelengths the human visual field can perceive range from 500nm-700nm.
Natural illuminant changes cause a gradual process the visual system can adjust to slowly, constantly compensating for changes in light therefore we continue to perceive an objects apparent colour (eg. Colouring of shadows as the day goes by against an object); this is known as successive colour constancy (Arnold, 2009). In contrast, when an illumination of an object changes rapidly over time, simple processes are not able to compensate for changes in illuminant colour therefore high-order processing steps in to enable our perception of an objects colour to remain appropriately consistent with the objects apparent colour; namely our visual system compensates for the light and provides an object colour that is close to its apparent colour (Arnold, 2009). This high-order processing enables us to perceive object colour relative to its apparent colour when light illumination changes instantly (eg. <1 second) although the colour of the object is tinted in the illuminant colour we can conceptualise its apparent colour through its object colour (Arnold, 2009). Simultaneous colour constancy occurs when 2 or more illuminants are lighting an object, although one half of an object may be shaded in green and another in red our visual system compensates for the illuminants to perceive a relatively consistent object colour. These processes have been described by various models that explain how the visual system computes this problem (Filipe & Alexandre, 2013).
Most models of colour constancy are grounded on the principle that cones responses corresponding to an objects reflectant colour are subject to transformational change as the stimuli relays to the visual cortex (Filipe & Alexandre, 2013). For example the colour of an object is detected by cones and a second stage process combines the contrast of colours therefore stabilizing the colour appearance across natural and rapid illuminant changes (pp 256, Goldstein, 2010). It is well known that for full colour vision (trichromacy) only three colour pigmants are required (L, M and S cones). Colour matching studies involve participants to adjust colour of a test object to match that of a control illuminated object; findings suggest that this is possible with any three colours (Arnold, 2009). Additionally further research has found that alongside primary coloured cones there is a visual system for opponent colour coding, i.e chromatic red/green mechanisms whereby the visual system discerns the levels of differences of colour between cones rather than what stimulus the cones actually received (Hurvich & Jameson, 1955).
Various fMRI studies have enabled researchers to recognise a visual pathway that extends to cortical modules that distinguish between visual frequencies (Arnold, 2009). A receptive field is mapped by stimulating a particular area in the retina with light and recording the response of the neurone in the visual cortex (Wandell, 1995). Stimuli from the retina converge onto a single retinal ganglion cell (rods: 120, cones: 6) in the area Striata (V1). Photoreceptors are rods and cones; rods function at low light levels and are responsible for monochromic night vision . Cones function at higher light levels and are responsible for high acuity colour daylight vision. Retinal ganglion cells are as follows; Magnocellular ganglion cells (M cells) carry information about movement, location and depth. Smaller Parvocellular ganglion cells (P cells) transmit signals for colour, form and texture of objects in the visual field. Koniocellular pathway cells (K cells) are suggested to be involved in colour vision. These stimuli are relayed to the lateral geniculate nucleus (LGN) in the Thalamus (Wandell, 1995). The LGN controls for various cognitive processes from visual stimuli for example spatial frequency selection and colour selectivity (Brainard, 2009).
Returning to the area Striata (V1), this module contains small receptive fields that are sensitive to changes in wavelength composition of light and are able to process successive colour constancy (Arnold, 2009). The light signals further travel down to the V2, an area located in the extrastriate cortex that processes simple properties inclusive of colour, motion and spatial frequency, this area also consists of opponent colour cells (Arnold, 2009). Area V4 activates when the receptive field receives multi-coloured stimuli, this area is concerned with automatic colour constancy operations regardless of memory of the colour of objects, and therefore it can detect colours of items the visual system has not perceived before (Zeki & Marini, 1998).
So do how do we know this is true? Researchers have been attempting to understand the functional properties of colour constancy since Newtonian times. The next half of this paper will discuss some of the models and studies performed in the past decade that has enabled scholars to understand these mechanisms.
To determine illuminants our retina field can distinguish researchers at Standford University measured the spectral sensitivity of cones in the retina of a Cynomolgus (Macaque) monkey (Baylor, Nunn & Schnapf, 1987). Baylor (et al, 1987) dissected the retina of the monkey to extract the cones; once extracted they shined light into the pipette containing the cone and recorded electrical current for each cone class and wavelength they were aware the visual system could detect (Baylor, et al, 1987). They tested all three cone classes; five L cones, twenty M cones and sixteen S cones. The spectral sensitivity of the cones were were logged and recordings found all three had a similar shape on a histogram chart but with different peak frequencies (Baylor et al, 1987). The photoreceptor proteins in the cones isomerize with photons of light independent of its wavelength; the response from the cone depends on the isomerization and corresponds with the wavelengths height of the curve on the histogram (Baylor et al, 1987). Their study provided evidence that cones do not have subclasses of colour perception therefore the three known classes are the only cones that receive visual stimuli (Arnold, 2009). They detected that changing the levels of photons of monochromatic light changes the relative response of the three cones, lights can be physically different but cause the same excitation in the cones. The cones are selective for different wavelengths of light. Additionally their results were congruent with results from psychophysical studies on colour matching as previously described (Baylor, et al, 1987).
So cones in the retina have different wavelength peaks and absorb photons of light regardless of its wavelength. We are able to determine object colour because each cone has an individual spectral sensitivity. This process is combined with the colour opponent system. Colour opponency is the process of the photoreceptors forming three opposing pairs (blue-yellow, red-green, black-white). For example the colour purple is redish and blueish yet colours are never greenish and reddish or yellowish and blueish. James and Hurvich (1975) attempted to understand how we can process these light frequencies with their hue cancellation study (Thompson, Palacios & Varela, 1998). Participants adjusted illuminants of mixtures of colours for example a mix of red and green lights until the light appeared neither red or green and resulted in appearing yellow. Therefore they support previous theories that there are four primary colours our visual system detects not just three (Thompson et al, 1998).
Supporting this theory De Valois and others (De Valois, Abramov & Jacobs, 1966) used microelectrodes to monitor individual neurones in the visual pathway. They found that excitation of one type of colour (+red -green) inhibits the perception of another colour (+green –red). A colour looks reddish when the red/green mechanism gives a positive response, greenish when the red/green mechanism gives a negative response (De Valois et al, 1966). Results of the study determined that the members of each pair cannot be detected in the same location which therefore explains why we cannot see blue-yellowish colours. This enabled researchers to understand why those with colour vision deficiencies such as seen in dichromats (only have two cones) confuse the colour of red with green or blue with yellow (depending on the missing cone). The simultaneous stimulation of L cones and M cones causes the inhibition of the colour blue+yellow enabling the perception of yellow (Brainard, 2009). Colour opponency is a higher tier process in the V2 that recodes L, M and S cones into their respective pairs once the photoreceptors of the retina absorb light.
Colour spectral sensitivity predicts receptor scaling. As mentioned earlier the histograms charted show a linear trend and similar curve for all cone classes in the Macaque monkey. Researchers speculated that long wave illuminants projected on the retina excite long wave cones (L cones) just as short wave excite short wave cones (S cones). To test this hypothesis Brainard and Wandell (1992) used natural light stimuli to illuminate different coloured rectangles on a CRT monitor that simulated matte surface (Arnold, 2009). Participants set assymetric colour matches between a control rectangle and various illuminated rectangles with differing spectral power distributions. They determined that for any change in illumination the mapping of test and control rectangles is a diagonal linear transformation (Brainard & Wandell, 1992). Therefore colour constancy is in part achieved due to adaptation of cones sensitivity to illumination change.
Evidence for higher processing of colour constancy in the perceptual field comes from research performed by Arend and Reeves (1986). As discussed earlier there are apparent colour and object colour when observing objects that are illuminated. Arend and Reeves showed that the two types of colour can be evoked by giving appropriate instructions to an observer (Arnold, 2009). They performed a study whereby participants were given two images; a test image (object colour) and an illuminated sheet of paper (apparent colour). They were asked to either modify an illumination so that it matched the test surface or the sheet of paper. They discovered that that when participants were asked to make the apparent colour match there was 20% colour constancy, conversely there was 78% colour constancy for object colour matching (Arend & Reeves, 1986). Therefore they suggested that the visual system can compensate for varied illuminated light and that higher level perceptual processes may be involved in determining what we perceive an object colour is with differing illuminants (Arnold, 2009). Researchers have suggested a combination of colour opponency, perception and wavelength frequency detection are all significant in achieving colour constancy (Zeki & Marini, 1998; Arend & Reeves, 1986).
In conclusion colour constancy is a difficult process the human visual system has to achieve. The retina visual field uses three cones to detect different wavelengths of light and each cone responds to a specific range of frequencies dependent on its own photoreceptor. Light reflecting off objects that differs to the colour of the object must be compensated for in the visual field so our visual cortex can discern an objects true colour regardless of any memory of that objects colour. Rapid changes in light require higher tier functioning between optic nerve and striatal cortex. Evidence has shown that alongside the retinal field containing cones that are relative to frequencies of light there are is also another process known as colour opponency that recodes frequencies received to enable the visual cortex to discern colours the cones are unable to process alone such as Yellow. Despite changes in illumination a normal visual system tends to perform well in perceiving an objects colour.