3 replies to this topic

[jessz]

I understand that cubic convolution is the best resampling method to use for downsampling data, but Photoshop doesn't have that option. Can I do this in Global Mapper, or is there some other (hopefully free) way to do it? Also, is there any rule I need to follow, such as reducing a file by set amounts (1/2, 1/4, etc). One final question: is there any elevation data for the U.S. that is in between ASTER and SRTM in resolution?

[David Medeiros]

I understand that cubic convolution is the best resampling method to use for downsampling data, but Photoshop doesn't have that option. Can I do this in Global Mapper, or is there some other (hopefully free) way to do it? Also, is there any rule I need to follow, such as reducing a file by set amounts (1/2, 1/4, etc). One final question: is there any elevation data for the U.S. that is in between ASTER and SRTM in resolution?

Not sure about the first set of questions. I usually use Natural Scene Designer to down sample and typically stick to commonly used resolutions (30m 90m 250m 600m 1km etc).

The USGS Seamless Server had NED data at several resolutions. 10m, 30m & 90m I believe.

[frax]

Photoshop has that option, it is called Bicubic there (it offers either sharper or smoother, two variants). Which option you use is also related to the nature of the data. With cubic convolution you want to watch coastlines, peaks and sinks, since you can get strange effects at the edges and where data changes in range. Best is to experiment!

I would actually suggest starting with nearest neighbour and see how it goes from there.

[Earthscientist]

I understand that cubic convolution is the best resampling method to use for downsampling data, but Photoshop doesn't have that option. Can I do this in Global Mapper, or is there some other (hopefully free) way to do it? Also, is there any rule I need to follow, such as reducing a file by set amounts (1/2, 1/4, etc). One final question: is there any elevation data for the U.S. that is in between ASTER and SRTM in resolution?

The answer to your question rather depends on what you want to do with the elevation data afterwards.

Any interpolation algorithm, other than nearest neighbour, used for up- or down-sampling or geomtrically transforming an arrayed dataset, will introduce point/cell/or pixel values into the resampled datset that may not have been present in the unsampled dataset and will also introduce artifacts. For work such as classification maintaining this can be very important and so use of anything other than NN should be avoided or at least minimised.

However, if the resampled dataset is only going to be used in visualisations then bilinear or bicubic convolution is recommended.