[gdal-dev] gdalwarp and NOAA L1b

[Andrew Brooks]

On Thu, 14 May 2009 20:21:11 +0100, yehya imam <yehya_imam at yahoo.com> wrote:
>
> I've downloaded AVHRR L1b images from NOAA CLASS. I'm trying to rectify the images via gdalwarp -tps. The output is obviously erroneous as per gdalinfo 

The error message "There is a problem to invert the interpolation matrix" explains why gdalinfo gave erroneous information.

I too have had great difficulties with gdalwarp and L1B.

The way the driver reports the geolocation data is not optimal and the methods used to process the data are also not optimal so it's no surprise that the result is often broken.

I submitted a patch to fix one of the issues but it was never applied; you can download it from here: http://trac.osgeo.org/gdal/ticket/2501

I also submitted some code to perform proper interpolation but nobody has made use of it.

I started a dialogue with the driver author who said he was going to rewrite it but then he went silent and I've not heard from him since.

One method I've used in the past to get some results is as follows:

1) Extract ALL of the geolocation data myself (51 points per scan line) since the driver normally subsamples
2) Interpolate this myself (Lagrangian method) to get 2048 points per scan line
3) Create two VRT files for the latitude and longitude data
4) Create a VRT file describing the image data in the file which references the two geolocation VRT files
5) Use the VRT file with my patched version of gdalwarp

This was successful for me in the past but when I tried it again recently it failed.  I can give you more details if you want though.

Andrew