Volume visualization for
large, potentially time-varying, datasets on high-resolution displays
(9Mpixels LCD panel, projection or LCD based tiled display, ...), refered
to as "Vol-a-Tile".
This document proposes to
report development and progress of the "Vol-a-Tile" system.
The following items are to
test-crash the pipeline from raw data file stored on disk to the display.
The usual steps are normalizing, converting to usual datatypes, downsampling,
storage, network transmission, and insertion into the 3D pipeline for
display.
Usability of the UI (transfer
function, ...) is also examined.
Only single-display image
are provided so far. Tiled-display iamges coming soon...
New screenshots on MacOSX
(July 10, 2004):
 |
"Bonzai" dataset |
 |
"Ranvier" dataset |
 |
"Arad" dataset |
Test datasets:
- Bonsai:
- Raw: 256 x 256 x 256
volume, "unsigned char" type, 16MB
- Comment: very easy
to classify, clear cut between the different components
- Arad:
- Raw: 1001 x 801 x
801 volume, "float" type, 2.5GB
- Processed: 512 x 512
x 512 volume, "unsigne char" type, 128MB
- Comments: large dataset
downscaled, normalized to [0..1], converted to "unsigned
char", and gaussian filtered to remove some noise (?).
- Jurassic tank:
- Raw: 440 x 290 x 148,
"unsigned char" type, 25MB
- Processed: 256 x 256
x 256 volume, "unsigned char" type, 16MB
- Comments:
- Earth shaking
- Raw: 512 x 512 x 512
volume, "unsigned char" type, 128MB
- Comments: Large datasets,
two areas in the derivative histrogram
- Visible Female:
- Raw: 512 x 512 x 1800
volume, "unsigned short" type, 950MB
- Processed: 512 x 512
x 512 volume, "unsigne char" type, 128MB
- Comment: not displayed
yet...
Relevant Links
|