[GRASSLIST:2110] Re: landsat processing

[Glynn Clements]

Ian Macmillan wrote:

> Hi all, I have a question about landsat processing.  i am trying to produce a
> nice 15m resolution "true color image" from landsat 7 data. I am using band 3
> (red) band [2+4]/2 (green) and band 1 (blue).  These three bands produce a very
> nice 30m res image with d.rgb.  I would like to incorporate the 15m res band 8
> though.  I am going about this by using i.rgb.his and i.his.rgb as described in
> the book.  I get an image that is 15m res, however the colors become a little
> distorted, namely, the urban areas become purple (instead of grey) and the
> vegetated areas become an ugly bright green.  I have tried using both an
> histogram optimized band 8 (r.rescale) and the original band 8.
> 
> Anybody have any experience with this?  I am really interested in eventually
> getting images that look like those produced by NASA here:
> 
> http://photojournal.jpl.nasa.gov/catalog/PIA03372
> 
> My next step will be to add shading using shade.rel.sh, or some modification of
> r.slope.aspect.  Any input would be helpful as I can't figure out how to keep
> the colors looking realistic.

One option which should preserve the colours would be to convert to
HSV with rgb.hsv.sh, replace or modify the intensity (V) channel, then
convert back to RGB with hsv.rgb.sh.

Note that neither of these scripts use or set the colour tables, but
d.rgb/r.composite require them be set correctly (you have to use
"r.colors color=rules" rather than "r.colors color=grey").

-- 
Glynn Clements <glynn.clements at virgin.net>