Previous: Naming Color Samples Up: Naming Color Samples
Complex color names
Once we have the sorted list of candidates
, we can not only return
the name of the best candidate category, but also a complex color name,
constructed from the names of the top two candidates. The following
algorithm presents a tentative way of doing this. I claim no significance
for this algorithm other than being an illustration of how one might go
about constructing compound names. I have not compared it with experimental
or other existing data on compound color names (if any).
- Given:
(equation
), but using
(equation
)
-
- the compound name is given by
, if
, if
, if
,
otherwise
where
selects the second element of a tuple, and
the third. The
lower threshold
is used to select both primary and secondary
candidates, because the secondary candidate's goodness value is allowed to
be lower than the regular threshold for category membership. Step 3.1
returns a null name (i.e. a no-category judgment) if no
candidates exceed the regular threshold
. Step 3.2 returns
only the primary candidate's name if no secondary candidate is found. Step
3.3 returns names of the form ``greenish blue'', where ``blue'' is
the primary name (highest membership value) and ``green'' the secondary, if
both candidates exceed the regular threshold
. Step 3.4,
finally, returns names of the form ``somewhat greenish blue'' if the
secondary candidate does not exceed
(but the primary candidate
does). Chapter
shows some examples of compound names
derived with this algorithm. Varying the parameters (magic numbers)
and
affects the behavior of the algorithm, but I have
not explored this in any detail.