I have to do a bilinear interpolation on an image with two no data values, using the utility gdalwarp of GDAL (http://www.gdal.org/gdalwarp.html), so these two values shouldn't be considered in interpolation. How can I do this? I evaluated the gdalwarp options "-srcnodata" and "-dstnodata", but they specify several no data values, with each value corresponding to one band. The case of my image is different, because I have only one band with multiple no data values.
If disk space and processing time aren't a major concern, then you could use gdal_calc.py to pre-process your data before gdalwarp. Create an intermediate file that has only one no-data value by converting one of the two no-data values to be the same as the other.
Suppose, for example, that your no-data values are 0 and 17. Then you might do something like this:
gdal_calc.py -A input.tif --outfile=result.tif --calc="A*(A!=17)"
Whenever the pixel value is 17, the expression A!=17 is 0, so those values get set to 0. Otherwise it's 1, so it'll leave the other values alone.
If 0 is not a no-data value in your dataset, you'll have to be a bit more clever with your calc expression.
I have recently came across the same issue. For example, in one raster data, I have two nodata values: 1.7976931348623157e+308 3.4028234663852886e+38
I used gdalwarp to set them to a unified nodata value: gdalwarp -srcnodata 3.4028234663852886e+38 -dstnodata 1.7976931348623157e+308 input.tif output.tif
Similarly, you can reset multiple nodata values (e.g. a, b, and c) to one unified nodata value (c) by running gdalwarp multiple times:
gdalwarp -srcnodata a -dstnodata c input.tif intermediate.tif gdalwarp -srcnodata b -dstnodata c intermediate.tif output.tif
Cheers.
