RFC 4: MapServer Raster Resampling

[Ed McNierney]

Folks -

I don't get a vote, but I think this is an excellent proposal.  My only
suggestion is that the "RESAMPLE=CUBIC" might be better named
"RESAMPLE=BICUBIC" as that it how I have always referred to that
algorithm, and how I've seen it described by others.

Thanks to Frank, Tydac AG and DM Solutions!

	- Ed

Ed McNierney
President and Chief Mapmaker
TopoZone.com / Maps a la carte, Inc.
73 Princeton Street, Suite 305
North Chelmsford, MA  01863
ed at topozone.com
(978) 251-4242   

-----Original Message-----
From: UMN MapServer Developers List [mailto:MAPSERVER-DEV at LISTS.UMN.EDU]
On Behalf Of Frank Warmerdam
Sent: Friday, September 16, 2005 1:14 PM
To: MAPSERVER-DEV at LISTS.UMN.EDU
Subject: [UMN_MAPSERVER-DEV] RFC 4: MapServer Raster Resampling

=====================================
MS RFC 4: MapServer Raster Resampling
=====================================

:Date:  2005/09/16
:Author: Frank Warmerdam
:Contact: warmerdam at pobox.com
:Last Edited: $Date: 2005/09/16 $
:Status: proposed

Overview
--------

Additional resampling kernels will be implemented in the MapServer GDAL
raster resampler code including "averaging", "bilinear" and "cubic". 


Technical Details
-----------------

 o The new resampling options will be implemented in mapresample.c and
will
   only be accessable for datasets processing through that mechanism
(ie. 
   GDAL raster formats, requires coordinate system information defined).


 o The resampling kernel to use will be selected with one of the
following.
   The default will be RESAMPLE=NEAREST, the current behavior. 
     PROCESSING "RESAMPLE=NEAREST"
     PROCESSING "RESAMPLE=AVERAGE"
     PROCESSING "RESAMPLE=BILINEAR"
     PROCESSING "RESAMPLE=CUBIC"

 o The mapraster.c code currently decides whether to invoke the "simple"

   GDAL renderer or the "resampling" GDAL renderer based on whether the
   projection seems to differ.  It will also now check for the RESAMPLE 
   processing option, and force use of mapresample.c if the resampling 
   kernel select is other than NEAREST.

 o Note that resampling kernels other than NEAREST can have a
substantial 
   effect on rendering performance.   For this reason NEAREST will
remain
   the default. 


Mapfile Implications
--------------------

All new options are selected via new PROCESSING options.   There is
no change in the mapfile syntax.  There should be no compatbability
problems with old mapfiles. 


MapScript Implications
----------------------

There are no additions or changes to the mapscript API.  The new options
are controlled via PROCESSING information on the layers which I believe
is already manipulatable from mapscript. 


Documentation Implications
--------------------------

The new processing options will need to be documented in the Raster
HOWTO (and possibly the mapfile reference). 


Test Plan
---------

New test cases for each mode will be incorporated in msautotest/gdal. 


Staffing / Timeline
-------------------

The new feature will be implemented by Frank Warmerdam and completed
by October 15th in time for the MapServer 4.8 release.   Implementation
is generously funded by Tydac AG and managed by DM Solutions. 


-- 
---------------------------------------+--------------------------------
---------------------------------------+------
I set the clouds in motion - turn up   | Frank Warmerdam,
warmerdam at pobox.com
light and sound - activate the windows | http://pobox.com/~warmerdam
and watch the world go round - Rush    | Geospatial Programmer for Rent