[gdal-dev] Nodata and compression artifacts

[Even Rouault]

> I often have multiple input files and didn't
> see a way to use the cutline option with multiple input cutline.

You can try merging them into a single OGR layer (with ogr2ogr for example)

> On Nov 25, 2013 3:33 PM, "Even Rouault" <even.rouault at mines-paris.org>
> 
> wrote:
> > Le lundi 25 novembre 2013 22:29:00, Simon Shak a écrit :
> > > I’m working with gdalwarp to reprocess a large amount of imagery to be
> > > compatible with another program that requires imagery to be in WGS84.
> >  
> >  The
> >  
> > > input imagery is compressed in MrSID format and does not include an
> > > internal mask for nodata.  I don’t know if this is because the creator
> > > of the imagery overlooked it, or if the format doesn’t support a mask.
> >  
> >  Either
> >  
> > > way, when I attempt to merge neighboring sets, I get odd bands of dark
> > > color.  I’ve looked closely, and it is evident because at the edge of
> > > the images are non 100% black pixels, that though I’m sending
> > > –srcnodata 0
> > 
> > into
> > 
> > > gdalwarp, they get read as pixels and progress through.  I’ve looked
> > > into using the nearblack command on the files first, but the
> > > compression ratio of the .SID files makes it such that the files don’t
> > > easily fit into my hard drive array for pre-nearblacking them before
> > > processing, plus the physical size of some of these files are large
> > > enough that the nearblack takes a long time to run.  Without the
> > > nearblack step, my multithreaded control script can process one chunk
> > > in a day, but adding the nearblack, and it increases to a week at
> > > least.
> > > 
> > > 
> > > 
> > > I’m looking for a solution that would not require making a large
> > > interim uncompressed version and would hopefully not incur a lengthy
> > > additional process.
> > > 
> > > 
> > > 
> > > The simpler thoughts I have would be to adjust gdalwarp’s –srcnodata to
> > > take a range option, much like nearblack, so that if it detects a pixel
> > > (even in the middle) that is with the range specified would get
> > > ignored,
> > 
> > or
> > 
> > > a way to include an ancillary file that could contain a mask.  Either
> > 
> > would
> > 
> > > work for me, I have potential ways to quickly generate a mask for the
> > 
> > input
> > 
> > > files.  I’d think the mask could work much like .TIF can have a .TFW,
> > 
> > that
> > 
> > > a .MSK could be detected as well.
> > 
> > You can use the -cutline option of gdalwarp if you have the mask as a
> > shapefile
> > or another OGR datasource.
> > You could also use a VRT file to combine the MrSID imagery and add
> > another band
> > from TIF for example as the alpha/mask band.
> > 
> > 
> > --
> > Geospatial professional services
> > http://even.rouault.free.fr/services.html

-- 
Geospatial professional services
http://even.rouault.free.fr/services.html