Ambient Diagnostics - Ambient Intelligence for Scientific Discovery - page 230

Db to represent roughness or smoothness of the tongue, 2) crack index CI to
describe cracks on the tongue, 3) energy function for describing the roughness,
and 4) entropy function to represent the order of the texture on the tongue:
F =[a ∗ , b ∗ , Db , CI , energy , entropy]

(1)

5.1 Color Space Coordinates
The tongue project presents a unique challenge in the overall gross variance of the
data set. There are tongues from different genders, ages and ethnic groups, which
present unique diculties in creating texture algorithms that are applicable to
the entire data set. Minimal usage of parameters is key to developing a robust
algorithm. A color space is a way of numerically describing a color. This almost
always requires three numbers to accurately and succinctly describe all possible
colors, as it is trivial to describe simple colors yet not so easy to describe colors
like a faint red, in low-lighting. A simple example of a color space is that one
may attempt to describe a color in terms of the actual color (red, orange, etc),
its tint, or how deep the color is, and how much lighting there is on the color.
Computer displays and televisions combine three primary colors (red, green and
blue) in different proportions to form the different colors of the spectrum.
CIE 1976 L*a*b* is a color space that is an attempt to linearize the per-
ceptibility of color differences. The non-linear relations for L*, a*, and b* are
intended to mimic the logarithmic response of the eye, where L represents the
lightness, a is the Redness/Greeness and b is the Yellowness/Blueness. However,
the values from a digital camera are RBG-based. RGB values in a particular
set of primaries can be transformed to and from CIE XYZ via a 3x3 matrix
transform. To transform from RGB to XYZ (with D65 white point), the matrix
transform used is [27, 39, 37]:

X
Y
Z

0 . 412453 0 . 357580 0 . 180423 R

G
B

= ∗

0 . 212671 0 . 715160 0 . 072169 (2)

0 . 019334 0 . 119193 0 . 950227

The L*, a*, and b* values can be converted from CIE XYZ, where coloring
information is referred to the color of the white point of the system, subscript
n .Here X n , Y n and Z n are the tristimulus values of the reference white.

( Y n ) 3 − for Y n 116 ∗ 16 > 0.008856

L ∗ = (3)

Y
Y n

903 . 3 ∗ otherwise

( f X
X n

− f Y
Y n

)

a ∗ = 500 ∗ (4)

( f Y
Y n

− f Z
Z n

)

b ∗

= 200 ∗ (5)

where f ( t )= t 3

for t> 0.008856

(6)

16
116

7 . 787 ∗ t + otherwise

Since we tried to eliminate the lightness effect, we only used a* and b* as
the color feature vector.