Distributed Post-processing and Rendering for Large-Scale Scientific Simulations - Scientific Visualization: Uncertainty, Multifield, Biomedical, and Scalable Visualization - page 374

Fig. 31.10 Warping of a delayed remote image ( left ) to the actual viewer position ( right ) based on
only one depth image leads to unpleasant artifacts
Fig. 31.11 Hybrid rendering allows interactive analysis of complex geometry like the depicted
compressor tera-byte dataset on a powerwall ( left ) and on a display wall ( right ). The surface geom-
etry is rendered locally at high frame rates whereas isosurfaces are rendered remotely
transmits images and the corresponding depth values per pixel from highly complex
scenes to the local client. Finally, these images are composed to a result. Noguera
et al. [ 37 ] used a hybrid rendering approach for terrain visualization on mobile
devices. While the geometry of the terrain close to the viewer is locally rendered,
geometry near the horizon with lots of triangles is rendered remotely. Engel et al.
[ 11 ] combines local and remote rendering in medical applications. Their system
uses a low quality local volume rendering while the viewer is changing visualization
parameters. When the view manipulation stops, a high quality image is requested
from the remote system.
Another approach is to split geometry into context geometry with low triangle
counts and complex feature geometry. While the context geometry is rendered with
local resources, complex features, such as isosurfaces, are rendered utilizing high-
performance parallel remote hardware (see Fig. 31.11 ). This can be used in order to
guarantee minimal frame rates on the client side, which is required in virtual envi-
ronments. Because remote images are received with delays due to network transfer,
they do not fit exactly to the local perspective. Therefore, image warping techniques
can be used to re-adjust the remote images to the current view.