[Geotiff] RasterPixelIsArea versus RasterPixelIsPoint and listgeo -tfw

Pierre Soille Pierre.Soille at JRC.it
Fri Jan 13 14:12:23 EST 2006


Hello All,

My understanding is that with PixelIsArea raster space, the tie point
coordinate refers to the coordinate of the upper left corner of the
considered pixel while, with  PixelIsPoint, the tie point coordinate is
simply that of the considered point (or indeed centre of the pixel if a
pixel were to be considered at that coordinate...).  See extract from
the GeoTIFF specifications  at
http://www.remotesensing.org/geotiff/spec/geotiff2.5.html#2.5.2

I just further googled on this issue (which is keeping me busy for some
time now).  And  found the following at
http://64.233.183.104/search?q=cache:HMENn5TmH70J:support.esri.com/index.cfm%3Ffa%3Dknowledgebase.techarticles.articleShow%26d%3D19491+RasterPixelIsArea+or+RasterPixelIsPoint&hl=it&lr=&strip=1

GeoKey 1025 (GTRasterTypeGeoKey) dictates whether the reference
coordinate is the UL (*RasterPixelIsArea*, code 1) or center
(*RasterPixelIsPoint*, code 2) of the UL pixel. If this key is missing,
the default (as defined by the specification) is to be
*RasterPixelIsArea*, which is the UL of the UL pixel.

See also page 8, point 6. at
http://edcdaac.usgs.gov/landdaac/tools/modis/info/ReleaseNotes.pdf

see again page 8 at
http://www.3001data.com/3001website/main/downloads_files/rage/arccatalog_rdt_geotiff_ewug2002.pdf

My interpretation is in line with these latter documents but it is only
an *interpretation* at this stage and it is not shared by everyone.  It
is quite important though to get to clarify  this issue because
otherwise precise geocoding cannot be achieved (for instance, with my
interpretation,  the tfw files generated by listgeo -tfw
are not correct for the  RasterPixelIsPoint model).

I came accross this issue while processing NASA Landsat 7 archives
where, surprisingly,  RasterPixelIsPoint model is used
(but I based my processing of the imagery on tfw's generated using listgeo).


Hopefully some reader of this email will definitely clarify this issue.


Best regards,

Pierre
EC Joint Research Centre
Ispra





More information about the Geotiff mailing list