different at the cost of greater similarity of the short-wavelength
limbs. The reverse holds true for negative disparities. The same
pattern is observed for protanomals.
3.2. The effect of optical density on color vision
For any theoretical observer (possessing any combination of
peak sensitivity and optical density) viewing any modeled scene,
our cell-counting method yields a single number that is related
to the richness of his color vision. However, this number will also
be affected by the quantity of chromatic information potentially
available in the scene (clearly some scenes will give rise to a more
varied chromatic percept than others). Therefore, in order to summarize
the data for many scenes we normalize the cell count for
each theoretical observer to the cell count for a ‘‘normal’’ observer
viewing the scene. We define our normal observer as having peaks
of M and L at 531 nm and 561 nm respectively, and expressing
both pigments at an OD of 0.4. We can then express the cell count
of each theoretical observer as a fraction of the cell count of the
normal observer viewing the scene, and take the mean of this fraction
to summarize the data from many scenes. Fig. 4 shows this
summary for theoretical deuteranomals and protanomals viewing
the two sets of hyperspectral images. The default assumption of
a 2:1 scaling of L:L0 and M0
:M is made here – other combinations
of cone scaling are considered in the next section.