GeoTiff with overviews results in strange rendering effects.

[georgbachmann]

I have a geotiff that I want to render using Tilemill. It's a quite large area, so I need to create overviews for it. Otherwise the rendering performance is just very bad.

So now when I use the tiff without overviews, everything renders fine, but... it would take about 60 days to export the region that I want to export. Which is bad... If I render the same area with a tif that contains overviews, it takes about one day, which is ok. The only problem is that as soon as my tif has overviews in it, it starts to make strange edges on the tiles. It seems like data is somehow shifted or translated or something. I attached an image to better see it. I marked the most obvious regions.

I created the overviews using gdaladdo -r average --config COMPRESS_OVERVIEW LZW input.tif 4 8 16 32. I also tried bilinear resampling, but same effect. So I stopped trying others...

Would be very glad for a pointer in the right direction. Or maybe a system configuration where it's working. Maybe an other gdal-version? Or an other OS? Or an other node version? So if somebody could tell me a working config, I'd be super happy.

Here are some stats about my geotiff (different file (greayscale), but same effect):

Driver: GTiff/GeoTIFF
Files: raster.tif
Size is 89560, 71226
Coordinate System is:
PROJCS["WGS_1984_Web_Mercator_Auxiliary_Sphere",
    GEOGCS["GCS_WGS_1984",
        DATUM["D_WGS_1984",
            SPHEROID["WGS_1984",6378137.0,298.257223563]],
        PRIMEM["Greenwich",0.0],
        UNIT["Degree",0.017453292519943295]],
    PROJECTION["Mercator_Auxiliary_Sphere"],
    PARAMETER["False_Easting",0.0],
    PARAMETER["False_Northing",0.0],
    PARAMETER["Central_Meridian",0.0],
    PARAMETER["Standard_Parallel_1",0.0],
    PARAMETER["Auxiliary_Sphere_Type",0.0],
    UNIT["Meter",1.0],
    EXTENSION["PROJ4","+proj=merc +a=6378137 +b=6378137 +lat_ts=0.0 +lon_0=0.0 +x_0=0.0 +y_0=0 +k=1.0 +units=m +nadgrids=@null +wktext  +no_defs"]]
Origin = (1037360.000000000000000,6290420.000000000000000)
Pixel Size = (10.000000000000000,-10.000000000000000)
Metadata:
  AREA_OR_POINT=Area
Image Structure Metadata:
  COMPRESSION=LZW
  INTERLEAVE=BAND
Corner Coordinates:
Upper Left  ( 1037360.000, 6290420.000) (  9d19' 7.55"E, 49d 5'29.80"N)
Lower Left  ( 1037360.000, 5578160.000) (  9d19' 7.55"E, 44d43'26.05"N)
Upper Right ( 1932960.000, 6290420.000) ( 17d21'50.67"E, 49d 5'29.80"N)
Lower Right ( 1932960.000, 5578160.000) ( 17d21'50.67"E, 44d43'26.05"N)
Center      ( 1485160.000, 5934290.000) ( 13d20'29.11"E, 46d57' 8.24"N)
Band 1 Block=89560x1 Type=Byte, ColorInterp=Gray
  NoData Value=0
  Overviews: 22390x17807, 11195x8904, 5598x4452, 2799x2226, 1400x1113, 700x557

**Environment: **

• OS & Version: Ubuntu:18.04 (running on Docker)
• Browser & Version: Safari 14.0.2
• Node Version: 8.15.0
• GDAL on host: 2.1.2 (used to create the overviews)
• GDAL in tilemill container: 2.2.3

Installation Method
Docker

[csytsma]

Georg, if you can provide a link to download the sample geotiff (ideally as small an image as possible) and project setup, I can try rendering on my instance, to see if I get the same results. The underlying mapping engine is Mapnik, so I'm going to guess it's an issue with Mapnik, not something specific to Tilemill. You could use Tilemill XML creation functionality, then render directly with Mapnik. You might have better luck posting to a Mapnik support channel with those results.

…

On Apr 27, 2021, at 11:58 PM, Georg Bachmann ***@***.***> wrote: I have a geotiff that I want to render using Tilemill. It's a quite large area, so I need to create overviews for it. Otherwise the rendering performance is just very bad. <https://user-images.githubusercontent.com/893113/116359492-24278180-a7ff-11eb-8250-c9bd29c99fdf.jpg> So now when I use the tiff without overviews, everything renders fine, but... it would take about 60 days to export the region that I want to export. Which is bad... If I render the same area with a tif that contains overviews, it takes about one day, which is ok. The only problem is that as soon as my tif has overviews in it, it starts to make strange edges on the tiles. It seems like data is somehow shifted or translated or something. I attached an image to better see it. I marked the most obvious regions. I created the overviews using gdaladdo -r average --config COMPRESS_OVERVIEW LZW input.tif 4 8 16 32. I also tried bilinear resampling, but same effect. So I stopped trying others... Would be very glad for a pointer in the right direction. Or maybe a system configuration where it's working. Maybe an other gdal-version? Or an other OS? Or an other node version? So if somebody could tell me a working config, I'd be super happy. Here are some stats about my geotiff: Driver: GTiff/GeoTIFF Files: raster_3857.tif Size is 299379, 186825 Coordinate System is: PROJCS["WGS 84 / Pseudo-Mercator", GEOGCS["WGS 84", DATUM["WGS_1984", SPHEROID["WGS 84",6378137,298.257223563, AUTHORITY["EPSG","7030"]], AUTHORITY["EPSG","6326"]], PRIMEM["Greenwich",0, AUTHORITY["EPSG","8901"]], UNIT["degree",0.0174532925199433, AUTHORITY["EPSG","9122"]], AUTHORITY["EPSG","4326"]], PROJECTION["Mercator_1SP"], PARAMETER["central_meridian",0], PARAMETER["scale_factor",1], PARAMETER["false_easting",0], PARAMETER["false_northing",0], UNIT["metre",1, AUTHORITY["EPSG","9001"]], AXIS["X",EAST], AXIS["Y",NORTH], EXTENSION["PROJ4","+proj=merc +a=6378137 +b=6378137 +lat_ts=0.0 +lon_0=0.0 +x_0=0.0 +y_0=0 +k=1.0 +units=m ***@***.*** +wktext +no_defs"], AUTHORITY["EPSG","3857"]] Origin = (649560.194258583476767,6085001.829428507946432) Pixel Size = (1.825432137611944,-1.825432137611944) Metadata: AREA_OR_POINT=Area Image Structure Metadata: COMPRESSION=LZW INTERLEAVE=PIXEL Corner Coordinates: Upper Left ( 649560.194, 6085001.829) ( 5d50' 6.35"E, 47d52' 6.52"N) Lower Left ( 649560.194, 5743965.470) ( 5d50' 6.35"E, 45d46'20.97"N) Upper Right ( 1196056.242, 6085001.829) ( 10d44'39.68"E, 47d52' 6.52"N) Lower Right ( 1196056.242, 5743965.470) ( 10d44'39.68"E, 45d46'20.97"N) Center ( 922808.218, 5914483.650) ( 8d17'23.02"E, 46d49'50.52"N) Band 1 Block=299379x1 Type=Byte, ColorInterp=Red NoData Value=255 Band 2 Block=299379x1 Type=Byte, ColorInterp=Green NoData Value=255 Band 3 Block=299379x1 Type=Byte, ColorInterp=Blue NoData Value=255 **Environment: ** OS & Version: Ubuntu:18.04 (running on Docker) Browser & Version: Safari 14.0.2 Node Version: 8.15.0 Installation Method Docker — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#2741>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAA7ORBHLBSXCBAWO7GACE3TK6WYXANCNFSM43WOCMJA>.