[Liblas-devel] LAZ / LAS extraction performance

[Martin Isenburg]

hi michael,

very nice presentation and beautifully explained. check this out everybody:
http://prezi.com/manlkjwyomj8/mg4-lidar-compression/  i think i know fully
understand the compressor since i worked extensively with 1D wavelet
transforms on audio coding before. looks like you are doing a haar transform
with a subsample/difference instead of an average/difference filter and
therefore you remain in the "integer domain" with all coefficients. neat.
since you need power-of-two windows (1024? 2048?) for this to work, i assume
you do some padding with zeros or so for the last window. the bit-plane
example seems to overlook the minus/plus sign of the coefficients. i assume
that this done with just another bitplane?

Does LASZip encode the integer coefficients or the calculated floating point
> values?  If the former, does it preserve most of the LAS structure but only
> replace the PDR payload?
>

the LAS header remains uncompressed and laszip adds one VLR. only the point
data records are compressed. the coder operates directly on the integer
coefficients. thanks for answering A, B, C, D ... people (or maybe only me)
were not sure how lossless LT really was since one often associates
(imperceptible) loss with wavelet coding.


> Do you see a relationship between claimed resolution and the relative
> efficiency of the two compression schemes?
>

i did as long as i was not sure about the losslessness. now not so much
anymore. it does not have to do with the gps time because (except
autzen-colorized-1.2-3.las) you only tested data sets without gps time.

i am in particular interested in what happened for 2398_400.las. does your
wavelet coder handle RGB exceptionally well? does laszip handles RGB
exceptionally poor? autzen-colorized-1.2-3.las also has RGB and the numbers
look completely different again. but here we also have gpstime playing a big
role. have you seen other examples that you can share where SID is that much
better than LAZ?

can you run LT compressor on the file without colors? it seems laszip takes
an ueber-proportional hit on the rgb coloring of this data set.

01/17/2011  06:53 PM         8,906,275 2398_400.laz
02/05/2011  08:56 AM         4,615,538 2398_400_no_rgb.laz

you can download it here:

http://www.cs.unc.edu/~isenburg/lastools/download/2398_400_no_rgb.laz

martin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.osgeo.org/pipermail/liblas-devel/attachments/20110205/96344b67/attachment.html