On 10/05/2013 05:21 PM, Frank Warmerdam wrote:
On Sat, Oct 5, 2013 at 3:31 AM, Rahkonen Jukka
<[email protected]> wrote:
Hi,
Have also definitely a try with JPEG compressed tiffs. You can save 90% of the
disk space with nominal effect on speed and quality, but that depends on your
hardware and imagery. It is still worth testing. The command to use is
gdal_translate -of GTiff -co tiled=yes -co compress=JPEG -co photometric=YCBCR
For overviews use
gdaladdo -r average -ro --config COMPRESS_OVERVIEW JPEG --config
PHOTOMETRIC_OVERVIEW YCBCR
With or without -ro is a matter of taste. By using internal overviews you will
save a bit more space because external overviews have some ballast but if you
want to archive images sometimes you can discard the external overviews and
save some disk space by that time.
Tiffs are not as hopeless than plain JPEGs of PNGs and images do not need to be
read totally into memory.
Untiled tiffs are usually written in stripes so for reading a box from an
arbitrary place inside the image it is enough to read all the scan lines which
intersect that box. It is still more data than what is needed with tiled images.
There are also clever tiff readers which can read fast, let's say, every 10th
row and column from uncompressed tiff and create an subsampled, nearest
neighbour subsample in-the-fly. I have used one such in the 90's and Intel 386
computers could handle pretty well uncompressed tiffs of size 10000x10000
pizels without overviews. I think that GDAL can't do that kind of selecting and
this is more nice-to-know stuff from the stone age. What to remember is that
tiles and overviews are essential for GDAL and Mapserver.
Folks,
Great advice from Jukka, and I'd just note that GDAL
*will* efficiently pull out reduced resolution images on
the fly without overviews if you have a file organized
with a one scanline block even in the year 2013. For
map serving you are still much better off for several
reasons with the tiles+overviews approach, but I do
love the properly that a one-block-scanline file can be
quickly subsamled by viewer applications like OpenEV
and QGIS.
Best regards,
Hi,
This is why the legendary BIL format is still used by many :)
Cheers,
Angelos
--
Angelos Tzotsos
Remote Sensing Laboratory
National Technical University of Athens
http://users.ntua.gr/tzotsos
_______________________________________________
mapserver-users mailing list
[email protected]
http://lists.osgeo.org/mailman/listinfo/mapserver-users
