Scalable Devices - Scientific Visualization: Uncertainty, Multifield, Biomedical, and Scalable Visualization - page 347

by warping the server's data to the correct viewing parameters. To compensate for
the missing data due to parallax either multiple depth peeled images are transmitted
or the missing information is requested as a small package from the server.
Diepstraten et al. [ 15 ] follow a different approach, instead of reconstructing the
full image from data received from the server, they propose to focus only on feature
lines. These lines can be transmitted efficiently to the client and rendered even on
the low end mobile hardware interactively.

29.3 Large Displays

Using large output displays has been of interest in visualization for a long time.
This is illustrated, for example, by early systems such as the PowerWall [ 62 ], or
sophisticated immersive virtual reality setups such as the CAVE [ 13 ].
In recent years, advances in digital cameras and computational photography
have created a lot of interest in creating, processing, and displaying high-resolution
imagery such as Gigapixel images [ 36 ]. Gigapixel images cannot be displayed on
regular monitors in their entirety at their original resolution, and thus are a natural fit
for the high display resolution of large-scale displays. One example is the Giga-stack
system [ 50 ], which targets tiled display walls and relies on the CGLX framework [ 16 ]
to display the correct image data on the individual LCDpanels comprising the display
array.
Similarly, recent advances and increased resolutions in scientific image acqui-
sition such as confocal microscopy, for example high-resolution biological image
stacks [ 34 ], or electron microscopy scans of brain tissue [ 33 ], result in image and
volume data of extremely high resolution, for which inspection on high-resolution
displays is of great interest. Figure 29.2 shows an example of a high-resolution brain
tissue block obtained via electron microscopy, rendered on a large display wall with a
total resolution of 13,660 × 3,072 pixels (40 megapixels), consisting of 40 individual

display panels.
A well-established approach to building large-scale displays is to use multiple
high-resolution projectors [ 56 , 57 ]. In such systems, projector calibration in order to
hide seams, brightness and contrast differences, and other artifacts is often a difficult
practical issue. Recent advances in LCD display technology have enabled building
large, tiled displays consisting of individual, relatively cheap LCD panels [ 50 ]. Such
display arrays are usually much cheaper and easier to maintain than projector-based
systems. Recently, Papadopoulos et al. [ 46 ] combined 416 LCD displays for the
1.5 gigapixel, 4-wall Reality Deck. A practical issue on LCD display arrays is the
bezel width of the screens used, which determines the width of the border between
the individual image tiles.
Apart from the practical issue of building and setting up the actual hardware of
large displays, we are most of all concerned with the necessary software infrastruc-
ture and visualization algorithms that enable efficient rendering on tiled displays.
Many visualization systems build on an underlying software framework—or mid-