Martin Isenburg wrote:
> Hello,
>
> check out today's press release for some fantastic news. Esri and
> rapidlasso are announcing to join forces and together develop a LiDAR
> compressor for LAS 1.4 in open source avoiding unnecessary format
> fragmentation. Their new “LASeasy†tool not only compresses but
_Very_ nice!
I particularly liked the caption for the illustration: "reordering
points along an adaptive space-filling curve".
I strongly suspect Hilbert is involved in this reordering. :-)
If I had to design it I would use a basic (square) block size
sufficiently large that it can be compressed/decompressed with
near-optimal efficiency, while avoiding the need to load lots of unused
surrounding terrain.
With ~10 ms to access a random disk block and 100 MB/s transfer speed,
about 1 MB of data is needed for good local balance, while an
internet-based structure would prefer somewhat smaller blocks, maybe all
the way down in the 10-100 KB range?
This would decompress into maybe 5-50K points, so ~100x100 m with common
point densities.
These blocks would then be ordered according to their bit-interleaved
(x,y) coordinates.
Am I close?
Terje
--
- <
Terje.M...@tmsw.no>
"almost all programming can be viewed as an exercise in caching"