Door Identification -- Color Issues

Eyal Amir and Pedrito Maynard-Reid II

Color Spaces

Color and its perception is a science in its own right. Many books have been written about the nature of color and several systems have been formed, most of which are decades old. We give a short description of the issues and options that must be understood to implement our color filtering. We refer the reader [Wyszecki and Styles 1982], [A. Berger-Schunn 1994], [Charles A. Poynton 1997] and [J. Bradley 1994] for further details.

The RGB color space is perhaps the most commonly used color description method. For example, most color monitors operate on RGB inputs. The RGB color space is a vector space with three dimensions. Each color is reporesented by a vector with the components specifying, respectively, the amount of red, the amount of green, and the amount of blue in the color. In most computer graphics systems, these values are represented as integers in the range 0-255, while other systems (such as the image processing routines in MATLAB) use the finer-grained continuous segment 0-1. 0 in both methods means "no signal" whereas the top value means "top signal".

A different color space is HSV, standing for Hue, Saturation, and Value, respectively. Hue describes the basic color in terms of its angular position on a 'color wheel'. If we consider the position of color on this wheel in terms of angles (MATLAB vision library functions consider take it to be in the interval 0-1), then the "primary" colors are represnted by the following values (we also write the RGB value to the right of the angle).


Saturation of a color is a measure of how "pure" the color is. Desaturated colors will appear washed-out, or pastel, whereas saturated colors will be bold and vibrant, the sort of colors you'd paint a sports car. In the RGB color space, you can desaturate colors by adding white to them. For example, if you take red (255,0,0) and add a medium grey (128,128,128) to it, you'll get a shade of pink (255,128,128). Note that the component values are clipped to remain in the range 0-255.

The third component of an HSV vector is Value which corresponds to intensity. It is a measure of how "bright" the color is. A color with Value 0 is as dark as possible (i.e., black).

Other color coordinate systems include CIE-LUV (L is Luminance, U is primarily blue (with some green information) and V is primarily red (with some yellow information)) and the CIE-LAB color space (L is luminance, A and B are chroma coordinates). These and other spaces are described in the references we mentioned above.

Eyal Amir and Pedrito Maynard-Reid II.
Last modified: Tue Mar 16 16:32:40 PST 1999