Information Technology Reference
In-Depth Information
3
The Use of Color Information in Stereo Vision
Nowadays, most digital cameras enable the acquisition of color images. The use of
color images becomes, therefore, more and more common in image processing and
computer vision. In this section, an investigation of the efficiency of color informa-
tion as an aid in solving the stereo correspondence problem is described.
3.1
Color Spaces
Color is an important attribute of visual information. Human color perception relies
upon three distinct types of photoreceptor cells in the retina, called cones. Similarly,
color images require three numbers per pixel position to represent color accurately.
The model chosen to represent color in a 3-D coordinate system is described as a
color space. The selection of an appropriate color space is of specific importance in
many applications, including stereo matching. The color spaces that are most used
can be distinguished into four categories [23]:
•
Primary systems: RGB and XYZ;
•
luminance-chrominance systems: Luv, Lab, YC
r
C
b
;
•
Perceptual system: HSV;
•
Statistical independent component systems: I
1
I
2
I
3
and H
1
H
2
H
3
.
3.1.1
Primary Color Systems
Primary systems rely on the use of the three primary colors : Red, Green and Blue.
The most common system in this category is the RGB space, as its three coordinates
are the reference colors in almost all the image acquisition processes. In this addi-
tive color space, a color image is represented with three numbers that indicate the
relative proportions of Red, Green and Blue. Any color can be created by combining
Red, Green and Blue values in varying proportions. The three components of this
primary system are highly correlated and dependent on the luminance information.
However, two colors may have the same chrominance, but a different luminance.
To consider only the chrominance information, coordinates can be normalized as
follows:
R
R
+
G
+
B
,
G
R
+
G
+
B
,
B
R
+
G
+
B
.
r
=
g
=
b
=
The advantage of this new color representation is that it is invariant to affine changes
in illumination intensity. In 1931, the international lighting commission (CIE) [24],
recommended the XYZ color system. While the RGB space is not able to represent
all the colors in the visible spectrum, any perceived color can be described mathe-
matically by the amounts of the three color primaries X, Y and Z. These coordinates
can be computed by using the following transformation matrix [5]
