[OSGeo-Discuss] Open File Formats andProprietaryAlgorithms

Bob Basques Bob.Basques at ci.stpaul.mn.us
Tue Aug 25 11:34:38 PDT 2009


Landon, 

Rather than sput off about something theoretical, I'll relate our use of
MRSID deliverables. 

This is our Typical publishing process for MRSID format data that we
receive, which happens fairly regularly. 

Our use case in this instance is for visual inspection of aerial
photography. 

* First we extract JPG tile pyramids from the MrSID format into
MapServer ready tile layers doubling/halving the resolution of the
output for each level of tiles along with generating WORLD files for
each tile. (we actually have a static grid that we use for tiling that
makes this a one time process, since the WORLD files are reusable across
data sets.) 
* Then, index the tiles for each level for Mapserver's use  We start at
Level "0" (L0 = 6" resolution) and increase by doubling resolution until
we have a single tile at the City view level, ~L7. 
* Mapserver is setup with a multi-level display with cutoffs for each
tiled layer at the appropriate scales. 
* This approach guarantees only four tiles at most (as long as the
client viewer pixel width isn't bigger than the tile pixel size of each
tile) are ever retrieved by MapServer to build a composite output image.


The first three steps could be automated into a package that ran
automatically for example from most MRSID deliverables. 

Another use case might be DEM analysis, which might require converting
into other formats as well as the data originating in some other format
besides MRSID. 

Still another use case might be spectral separation and handling the
consequent analysis tasks. 

Anyway, the list goes on and on I sppose. 

bobb 



>>> "Landon Blake" <lblake at ksninc.com> wrote:


Bob, 


What did you have in mind when you said “pre-processing”. 


Landon 



Office Phone Number: (209) 946-0268 



Cell Phone Number: (209) 992-0658 



  


  




From: 

discuss-bounces at lists.osgeo.org [mailto:discuss-bounces at lists.osgeo.org]

On Behalf Of 

Bob Basques

Sent: 

Monday, August 24, 2009 9:21 AM

To: 

OSGeo Discussions

Subject: 

RE: [OSGeo-Discuss] Open File Formats andProprietaryAlgorithms 



  


Landon, 



  


Just had another thought . . . 



  


What about setting up a (openSource) tool set specifically for handling
Raster images for pre-processing purposes.  Might even be something that
publishers could re-distribute with their datasets, as in this processor
stack works with our data. 



  


Just thinking on the run here, no detail (or ramifications though
through [at all  :c) ]) 



  


bobb 



  




>>> "Bob Basques" <Bob.Basques at ci.stpaul.mn.us> wrote: 


Landon, 



  


>It would be interesting to come up with a standard structure on a
computer file system that could be used to accessed tiled raster data,
if this hasn’t been done already. One the file system structure was
defined, it would be fairly easy to write open source software that
accessed this structure and provided individual tiles as a service to
desktop GIS applications.   



  


Hmm, interesting angle, to expand on your idea a bit more, what about a
processing suite (or set of suites) that process data for different
types of uses, visual display, DEM analysis, etc.  Each "processor"
stack would/could have it's own rules associated with data resolution vs
files sizes, etc.   



  


bobb 









>>> "Landon Blake" <lblake at ksninc.com> wrote: 


  


Bobb wrote: “Here's my reasoning, we're never (ever?) going to hit the
top end on how big files ever get, resolution just keeps going up and
up, so there is always going to be some upper limit that will need to be
breached somehow.  Working out a proper method for segregating the data
up front (dare I say it), as some sort of standard (which can be
adjusted as time passes) will make everything work nicely, then all will
work with available tools when they are available, if tools to handle
larger datasets become available, and the communityI agree with some of the points you are making in your argument Bobb.
There is certainly a practical limit to how much you data you should put
in a single file. That is why we have lumber cut to 8 foot lengths. You
don’t need a flatbed semi to carry it to your house. :] 


  


                


  


When you refer to a standard for splitting data up front, what do you
mean? 


  


                


  


It would be interesting to come up with a standard structure on a
computer file system that could be used to accessed tiled raster data,
if this hasn’t been done already. One the file system structure was
defined, it would be fairly easy to write open source software that
accessed this structure and provided individual tiles as a service to
desktop GIS applications. 


  


                


  


Landon 


  



  


From: 


discuss-bounces at lists.osgeo.org [mailto:discuss-bounces at lists.osgeo.org]



On Behalf Of 


Bob Basques 


Sent: 


Monday, August 24, 2009 7:33 AM 


To: 


OSGeo Discussions 



Subject: 


RE: [OSGeo-Discuss] Open File Formats and ProprietaryAlgorithms 


  


                


  


All, 


  


                


  


Ok, I'm probably going to get someone irritated, but here goes . . 


  


                


  


Why not approach this from the other end of the spectrum and work at
making the original files smaller.  Work with the providers to make the
images smaller in the first place, or at least come up with a maximum
practical size to work with, I mean if this is the only (or biggest
reason) for implementing JP2, then getting folks to make the smaller
deliverables seems like a better long term approach. 


  


                


  


Here's my reasoning, we're never (ever?) going to hit the top end on how
big files ever get, resolution just keeps going up and up, so there is
always going to be some upper limit that will need to be breached
somehow.  Working out a proper method for segregating the data up front
(dare I say it), as some sort of standard (which can be adjusted as time
passes) will make everything work nicely, then all will work with
available tools when they are available, if tools to handle larger
datasets become available, and the community feels there is a
reason/need that these new larger files need to be handled, then they
get to change the standard. 


  


                


  


bobb 


  









  




>>> "Fawcett, David" <David.Fawcett at state.mn.us> wrote: 


  



I realize that there are likely not a large number of people who have
the expertise and experience to write this kind of code. 

Is this a project that should be shopped around for funding?  Google
Summer of Code?  A grant from our ~benevolent overlord Google?  Some
other foundation or org interested in open data formats? 

David.
-----Original Message-----
From: discuss-bounces at lists.osgeo.org
[mailto:discuss-bounces at lists.osgeo.org] On Behalf Of Michael P. Gerlek
Sent: Thursday, August 20, 2009 4:36 PM
To: OSGeo Discussions
Subject: RE: [OSGeo-Discuss] Open File Formats and Proprietary
Algorithms
<snip>


> Do you know why there hasn't been a broader adoption of JP2?

Not through lack of trying on my part :-)

I think the two biggest reasons are:

(1) The algorithms for handling large images in memory really are rocket
science, and no one in the FOSS community has gotten the "itch"
sufficiently bad enough to go and do the work needed inside the existing
open source packages.  Hopefully someday someone will.


_______________________________________________
Discuss mailing list
Discuss at lists.osgeo.org http://lists.osgeo.org/mailman/listinfo/discuss
_______________________________________________
Discuss mailing list
Discuss at lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/discuss 


  


Warning: 



Information provided via electronic media is not guaranteedprohibited. If you have received this information in error, please
notify the sender immediately. 


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/discuss/attachments/20090825/3dd035f6/attachment-0002.html>


More information about the Discuss mailing list