DIVISION OF

Optometry

Colour Constancy

Background

In our everyday experience colours do not appear in isolation. Our visual system has evolved and is adapted to a world where colours belong to objects that are seen in a context full of shadows, inter-reflections, highlights and other objects (see the fruitbowls below). I believe that there is much to be learned from studying human visual perception with stimuli that incorporate this richness. Most current studies on colour perception and colour constancy are carried out using simplified laboratory stimuli: flat scenes generally simulated on a computer screen. How much we can infer from these studies depends on the validity of two assumptions: that colour and other visual attributes (shape, texture, motion) are independent from each other and that conclusions obtained using self-luminous displays reflect performance in the real world.

Picture (1) (top left) was taken under daylight illumination while picture (2) (bottom left) was under fluorescent light.

The colour appearance of the fruit in both images is very different. This is because the film and camera do not possess colour constancy. If instead of taking a picture you had just looked at the fruit bowl under the different illuminations you would not have been aware of the difference. This is because our visual system possesses colour constancy: it can, from the light reflected by the scene, separate surface reflectance (the commonly called 'colour' of the object) from the spectral composition of the scene illuminant (the 'colour' of the light). These images also helps to illustrate how objects appear in everyday life. Notice the highlights, mutual illumination and shading. (Images courtesy of Dr. A C Hurlbert, University of Newcastle).

Picture (3) on the left creates a compelling illusion of a 3D scene. In the context of this image it is possible to distinguish between a variation in illumination (shading) and a variation in surface reflectance (colour). The top left cube is non-uniformly illuminated, its left face is the brightest while the front face is dim. We are aware of the difference in brightness and yet we see this cube as having the same colour (or reflectance) on all its faces. Now look at the detail of the image reproduced below. It corresponds to the area surrounded by dotted lines.The effect is more compelling if you place a piece of paper with a cut-out on top of the large image so that you isolate the area surrounded by dotted lines.

In this image the distinction between illumination and reflectance (colour) cannot be made. But when the two greys are seen in the context of a complex image our visual system uses the information provided by the other elements of scene and assigns the same reflectance (colour) to both regions of the cube.

Aims

• To investigate how well the human visual system can distinguish a change in surface colour from a change in the illumination (see figure 2), using an environment with controlled light-sources, shadows, highlights and inter-reflections.
• To compare performance for different psychophysical tasks in full cue situations (binocular viewing, shadows, highlights and mutual illumination) with performance as I successively remove cues (monocular view, no-shadows, etc.) in order to determine the relative weights they have in visual perception, and how they influence the outcome in both high- and low- level tasks.
• To develop and compare how an 'ideal observer' (a mathematical idealisation) would perform the tasks under the same conditions as our human observers. This will aid in establishing which are the crucial factors of the scene taken into account by our human system and so contribute to developing a model that will allow us to make predictions of human performance in novel situations.

Investigators.

• Dr. Marina Bloj
• Dr. Alexa Ruppertsberg