[gdal-dev] trying to change regionator image resampling to nearest-neighbor

[Roger André]

Hi Mary Jo,

Sounds like you have started down a slightly different path to resolve
this by using gdal2tiles.py, but I'm still a bit curious about the
results you were seeing with the regionator.  Would you mind sending
me a couple screenshots that show the before and after images, with
haloing?  Your description makes me think that it is a jpg compression
problem, more than a resampling artifact.  It might be possible to
simply change the output format to PNG in the code, and eliminate the
problem.

Regards,

Roger
--



On Thu, Nov 12, 2009 at 11:32 AM, Mary Jo Brodzik <brodzik at nsidc.org> wrote:
>
> Dear gdal-dev list members,
>
> I'm using a python tool called the regionator
> (http://code.google.com/p/regionator/wiki/ Welcome) on a large (14352x7440)
> png image.  The regionator uses the GDAL library to resample subsets of the
> original image into much smaller chunk images.  These images are set up at
> increasing levels of detail in a kml file that make the download of this
> information into a googleEarth client much more efficient.
>
> The png is color and includes transparency; its approximate resolution is
> 500 meters. I basically have 4 colors in the original image, black, white,
> blue and yellow. The output images that regionator is producing are great,
> but when I zoom in to distances where the original pixel sizes are visible,
> it looks like the resampling did some averaging across the 4 colors, so I'm
> getting blue sections surrounded by a blue halo that bleeds into white, and
> yellow and blue values bleeding into each other, etc.
>
> I'd like to control the resampling, and force it to do nearest-neighbor
> rather than averaging to eliminate the halos.  It looks to me like the
> regionator is doing the resampling with calls to ReadRaster and WriteRaster
> in in a class called extractor.py, but I don't know enough about the GDAL
> libraries to know how to change them. Is it possible to force the resampling
> to be nearest neighbor?  extractor.py isn't terribly large or complicated,
> so I'm including it here (below).
>
> I've already tried the googleearth forum and the kml-developers forum, no
> help from those communities.  Python's new to me, and I've only used the
> GDAL command-line utilities up to now.
>
> Thanks in advance for any tips.
>
> Mary Jo
>
> Here's the regionator's extractor.py:
>
> """
> Copyright (C) 2006 Google Inc.
>
> Licensed under the Apache License, Version 2.0 (the "License");
> you may not use this file except in compliance with the License.
> You may obtain a copy of the License at
>
>     http://www.apache.org/licenses/LICENSE-2.0
>
> Unless required by applicable law or agreed to in writing, software
> distributed under the License is distributed on an "AS IS" BASIS,
> WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
> See the License for the specific language governing permissions and
> limitations under the License.
> """
>
> """
> $URL: http://regionator.googlecode.com/svn/trunk/kml/extractor.py $
> $Revision: 200 $
> $Date: 2007-01-12 09:12:48 -0700 (Fri, 12 Jan 2007) $
> """
>
> """ class Extractor
>
> Phase 3 of SuperOverlay creation.
>
> Extract tiles from an image
>
> Front-ends GDAL.  Input file and output driver must be
> some format known to GDAL.
>
> """
>
> import gdal
> import tempfile
> import os
>
>
> class Extractor:
>
>  """ class Extractor
>
>  Create for Extractor the image file.
>  Each tile is resampled to the specified size and written
>  to a file of the given format.
>
>  """
>
>  def __init__(self,imgfile,twid,tht,fmt,verbose=False):
>
>    """
>
>    Args:
>      imgfile: a gtiff
>      twid,tht: extracted tile resample pixel width/height
>      fmt: GDAL output driver name ('PNG','JPEG')
>    """
>
>    self.__in_ds = gdal.Open(imgfile)
>    self.__twid = twid
>    self.__tht = tht
>    self.__fmt = fmt
>    self.__verbose = verbose
>
>    self.__bands = self.__in_ds.RasterCount
>
>    if self.__verbose:
>      print 'Extractor %s %d bands' % (imgfile,self.__bands)
>
>    # Intermediate work must be in GTiff (?)
>    self.__gtiff_driver = gdal.GetDriverByName('GTiff')
>
>    # Output driver/format is whatever the user specifies
>    # XXX handle bad/wrong fmt's less gracelessly
>    self.__o_driver = gdal.GetDriverByName(fmt)
>
>  def Extract(self,x,y,wid,ht,basename):
>
>    """ Extract a tile into a file
>
>    The given tile is extracted, resampled and saved
>    according to the tile pixel dimensions and format
>    specified at object __init__().
>
>    Args:
>      x,y: pixel offset
>      wid,ht: pixel dimensions
>
>    Returns:
>      True: complete success
>      False: any failure
>
>    """
>
>    # Get the tile's pixels resampled
>    twid = self.__twid
>    tht = self.__tht
>    i_data = self.__in_ds.ReadRaster(x,y,wid,ht,buf_xsize=twid,buf_ysize=tht)
>
>    # Have to Create out to GTiff first (?)
>    (fd, tmpfile) = tempfile.mkstemp(suffix='GTiff')
>    os.close(fd)
>    o_ds = self.__gtiff_driver.Create(tmpfile,twid,tht,bands=self.__bands)
>    o_ds.WriteRaster(0,0,twid,tht,i_data)
>
>    # Save off using the specified driver
>    filename = '%s.%s' % (basename, self.__fmt)
>    options = []
>    if self.__fmt == 'JPEG':
>      options.append('QUALITY=90') # Default of 75 is low quality
>    self.__o_driver.CreateCopy(filename, o_ds, options=options)
>
>    if self.__verbose:
>      print filename
>
>    # Delete the tmp file
>    self.__gtiff_driver.Delete(tmpfile)
>
> _______________________________________________
> gdal-dev mailing list
> gdal-dev at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/gdal-dev
>