Heat maps provide a quick glance on the data distribution over the picture observed during an
experiment.
Traditionally, heat-maps visualize the background image with overlaying semi-transparent
colors.
Most such visualizations use red color for highlighting the most intensively observed areas, and blue
or black for coloring the unobserved parts of the picture.
Colors in a heat map change gradually, and the display resembles a topographic image with hills and valleys.
Presenting the unobserved areas in black can be regarded as shading the image.
Usually, the shadows do not totally hide the background image, but only dim it.
Our approach substantially extends the notion of heat
maps. Here, the term “heat map” pertains to all the
visualizations where eye gaze data adds transparency, or
some color, to the background image. In our current
implementation, the background can be partly or totally
hidden by a shadow or fog (Fig. 4).
The initial opaqueness of an image is adjustable.
Transparency can be presented in either the traditional way
(using shades of gray), or employing some color scheme.
In the latter case, color in a particular location conveys the
intensity of the observation similar to the transparency
level in a gray-scaled heat map.
The intensity is proportional to the duration of the
observation. Thus, longer fixations add more transparency
than shorter ones. A fixation longer than some threshold
makes the display totally transparent at the location it is
superimposed.
In fact, the measured gaze position during an
observation contains only a single point that covers only a
relatively small part of the display. Meanwhile, the actual
observation involves a whole group of pixels.
Therefore, we suggest that every pixel in the heat
map related to a particular gaze location “extends
transparency” to the neighboring area. The radius of this
area should be a few degrees of arc and subject to the
user’s choice. We also propose three alternative forms for
the function of the transparency distribution. One of these
is a simple linear relationship, whereas the other two are
nonlinear (a sum of linear and sine wave, and a Gaussian).