Abstract
For certain types of data visualizations where the cost of
producing images is high, and the relationship between the rendering
parameters and the images produced is not quite obvious, a visual
representation of the exploration process can make the process
more efficient and effective. Image graphs represent not only
the results but also the process of data visualization. Each
node in an image graph consists of an image and the corresponding
visualization parameters used to produce it. Each edge in the
graph shows the change in rendering parameters between the two
nodes it connects. The image shows an image graph as a result
of exploring a data set from a combustion engineering simulation
of an industrial furnace. The goal was to reveal the temperature
distribution inside the furnace. From this graph, one can deduce
that the user was first searching for an appropriate color transfer
function before deriving the desirable visualization shown in
the lower right image.
Image graphs thus help streamline the process of visual data
exploration in two ways. First, the graphs give the user a representation
of the relationship between visualization parameter changes and
the resulting images. Often, these relationships are not obvious
from inspection of the rendered images. An understanding of how
specific rendering parameter changes affect the image output
is important because it reduces the number of images the user
must produce to find parameters leading to useful images, images
that can be quite time-consuming to produce.
Second, image graphs are not just static representations:
users can interact with a graph to review a previous visualization
session or perform new rendering operations. Operations which
cause changes in rendering parameters can propagate through the
graph. This capability accelerates the search for good rendering
parameters. Other dynamic features of the graphs, such as annotation
and automatic pruning, facilitate collaboration and animation.
The image sequence below shows an example of the propagated
effect before and after manipulating two edges of an image graph
produced from an MRI data set. First, a "color" edge
is being detached from node 7, and re-attached to node 3. This
action will replace the color transfer function of node 3 with
the color map of node 7 and trigger a re-rendering at node 3.
Furthermore, the effect of using a new color transfer function
at node 3 will propagate through its peers.
Second, we show the resultant image graph after forward-propagation
of a new property, in this case, the color transfer function.
Compared to the images in the top graph, note that nodes 3, 4,
5, and 6 have all been updated. Node 7 has been removed since
it has become redundant to node 3.
Finally, we show the resulting image graph after forward-propagation
of a new property, in this case, rotation. Compared to the images
in the middle graph, except the upper-left image (node 0), all
other images are updated using the new viewing angle. The propagation
triggered thus allows the user to see all the visualization results
on the right from a different view angle, using node 1. (Note
that the user was able to create new visualization results without
introducing new graph nodes.)
A remote volume visualization system we have built uses this
image graph interface. The results of a user study we have conducted
using twelve scientists indicate that image graphs reduce the
average amount of time needed produce meaningful images of complex
volumetric data sets.
Publications
Contact
Kwan-Liu Ma
