The paramount growth of data in recent years is undeniable, out of which
approximately 60% is spatial data. Additioanlly most of these data have billions of
records, making them difficult to process. The need of inerteractive visualization
of these spatial data is hence unavaoidable.
A popular approach for interactive big spatial data visualization is through
multilevel image indexes which organizes fixed-sized tiles in a pyramidal structure
to produce multilevel visualization. This leads to an exponential growth in the the
number of tiles with growing levels. For example, a webmap with 18 zoom levels can
have 90 billion tiles. Another common approach is to create data index like R-tree,
R+ tree and generate images from them only on user requests. This method however
fails to serve big data in real time.
When a user visualizes these maps, they execute visualization query by clicking
on a position of the map that they want to visualize or by zooming in or out of a
particular area. The visualization server is repsonsible for generating these images
in real time so that the user enjoys absolute interactivity. The indexes discussed
above are responsible for providing the user with the desired interactivity.
The image indexes with all pregenerated image tiles are highly interactive, but
has a high preprocessing overhead. On the contrary, the data indexes have no such
downside. But they are designed to respond only to range queries, only when the
range is small enough to processed on-the-fly.
Our design meets the two design midway and provides an index that is capable of
having both data and image components in the same index at the same time. To top it,
we also have a controlling parameter which decides the type of each of the nodes
(data or image) in an index. In our first paper, this parameter is the size of the
individual nodes.