2. Color

Color is a prime example of machine perception that cannot be properly understood without knowing in what way the human brain perceives color. Color cannot be defined based on physical properties of the stimulus alone; the way in which the human visual system observes the electromagnetic energy (what we call light) is important as well.

In this chapter we will discuss color from a colorimetric point of view. Colorimetry is a science that is aware of the need for a psycho physical description of color but aims at reducing the role of the human observer to being a judge about color equivalence. Telling whether two colors are the same or not (in a controlled experimental setting) is a task that is remarkably consistently performed over many different observers with great reproducibility. Quantifying color differences or even naming colors is a task whose outcomes vary greatly over human observers.

Color is an important visual clue in the computer sciences nowadays. Using color in either ‘color generating’ applications (computer graphics and visualization) as well its use in ‘color analysis’ applications (computer vision and image processing) can only be based on an understanding of the basics of color science. This chapter aims at providing that basic background material.

We will briefly introduce the science of colorimetry. We will learn that colors are physically characterized by the light spectra, the human eye has three different receptors with different color sensitivities thereby effectively projecting an infinite dimensional space of color spectra onto a 3D dimensional space of color responses.

The color matching experiment then enables us to define three vectors in spectral space that serve as the canonical basis of three dimensional color space (\(XYZ\) color space).

The tristimulus color theory has the important consequence that many colors can be generated by mixing three primary colors. Not all colors can be made with three primaries and any color reproducing device (a printer or a beamer or a screen) can only reproduce the colors in its gamut defined by the three primary colors. These primaries may (and will) differ from device to device therefore the need arises to adjust colors so each device generates comparable colors.

Whereas color reproduction is relatively easy, color recording is much more troublesome. Color recording requires three sensitivity curves that are linear combinations of the eye sentivity curves. In such cases the camera is called colorimetric. Few camera’s are colorimetric.

Then we describe some well known colormodels and their relation with the \(XYZ\) color model.