The idea behind the iconographic technique developed by Pickett, Grinstein, and Levkowitz is to produce
textural representations of the data in order to take advantage of the capacity of the human visual system
to perceive textures to detect statistical structures in data. The way in which textural representations of
the data are produced from icons has been described in [28, 29, 12].
The basic approach is to transform each datum (or each record in a database) into a graphical object (icon),
the visible features of which are under data control. By displaying the icons en masse, structures in the data
are revealed as streaks, gradients, or islands of contrasting texture. This is a very promising idea, but there
are several difficulties to be faced in pursuing it. Some of these difficulties are related to the mechanism
used to convey information to the user (texture). Others are related to implementation of the required
display techniques.
The notion of visual texture persists as one of the most elusive concepts. Extensive research has already
been done to define models explaining mechanisms of texture discrimination. The work of Julesz
[22] and Haralick [19] are representative of the effort devoted to understanding the
mechanism of texture discrimination. Resnikoff [32] extended the work of Julesz by developing a
more general mathematical model for analyzing and synthesizing textures. Despite these efforts the exact
mechanism by which the human visual system is able to recognize and differentiate textures is still
unknown. Thus, even though structure in the data can be seem in textures differences, we still have the
problem of finding out exactly what has been revealed.
The mechanism used to create textural representations of the data in the iconographic technique also presents
some difficulties. Pinkney has pointed out and addressed some of them. It is difficult to determine the most
appropriate mapping to create an effective visualization because first, there are many possible mapping
permutations to choose from, and second, measuring the effectiveness of a particular mapping is a
demanding chore that requires considerable effort.
While the iconographic technique offers excellent global views of a data set, examining the real values associated
with specific portions of the visual representation is not easy with current implementations. Pinkney has
developed several interaction techniques that enhance the process of understanding the relationship
between patterns revealed in an iconographic visualization and the relation or trend in the underlying data they
represent. However this understanding is still more qualitative than quantitative, and the iconographic technique
would benefit if methods of quantitatively measuring relationships behind revealed patterns were
developed.