EDGE-EQUALIZED TREEMAPS - Information Visualisation: Techniques, Usability and Evaluation - page 111

2.2 Representations for hierarchical structures
Various representations have been derived from Treemaps. Squarified [3]
and Circular [4] Treemaps are characterized by the shape of their nodes. A
Squarified Treemap is a representation in which the node's shape is as
close as possible to a square, and a Circular Treemap is a representation
where nodes are drawn as circles instead of rectangles.
Ordered [5] and Spatially Ordered [6] Treemaps are characterized by
the positions of their nodes. An Ordered Treemap targets ordered
hierarchies, and its representation maintains the order. A Spatially Ordered
Treemap is a representation targeting hierarchies with geographical
location information, maintaining the location relationships among nodes.
Other representations exist, such as the Cluster Treemap [7], which is a
representation in which the distance between data is reflected by the
relative positions in the Treemap.
In the above-mentioned representations, the widths, heights, or radii of
leaf nodes are not equalized. Therefore, when we embed charts in leaf
rectangles, it is difficult to equalize the scale intervals of the charts across
the entire Treemap.
The original Treemap divides the drawing area recursively. Therefore,
there are no unused spaces, but the widths and heights of rectangles vary.
For some data, it would be helpful to equalize the shapes of rectangles,
and some such representations have been proposed. Quantum Treemaps
[8] represent nodes as rectangles that are proportional to the object size;
the aspect ratio is suitable for the representation of important photographs.
HeiankyoView [9] represents hierarchical structures through containment
relationships.
Schreck et al. [10] developed a method of representing branch nodes as
a grid, based on Quantum Treemaps. Because the widths and heights of the
leaf rectangles are all the same in this representation, it is easy to equalize
the scale intervals of the horizontal axes when we embed the charts.
The three representations discussed above have rectangles whose
shapes are the same, and the area of these rectangles cannot be adjusted
according to the data drawn on the chart. When a chart contains data with
very large values, the scale interval of the relevant axis becomes smaller. If
we equalize the scale intervals in existing variations of a Treemap, a chart
with a smaller amount of data cannot use its assigned area efficiently.