rescale - what is it doing?

[mdavid800]

HI Guys ,

Hopefully this isnt a daft question but i am trying to understand what Titiler is doing so i dont make a mistake.

I have a COG data ser of Wind speed and it is in Float32 . In order to display this correctly i need to use the rescale function . However , i was hoping someone could confirm my understanding is correct .

My data set ranges from 4.9m/s - 16.4m/s , when I use rescale am i correct in assuming that the data range min - max is split between 0-255 ? i.e. 0 represents 4.9m/s and 255 represents 16.4m/s ?

if the above is correct, is its possible to set the colormap such that the anything above say 10m/s is at the top of the colour scale ? I.e. from a wind energy perspective 16m/s is an extreme and i don't want my colour map scaled according to this outlier .

code for my get request is below.

r = httpx.get( f"{titiler_endpoint}/cog/tilejson.json", params = { "url": url, "rescale": f"{minv},{maxv}", "colormap_name": "viridis" } ).json()

[geospatial-jeff]

Rescale is doing a linear stretch, the minv and maxv parameters in your GET request determine the input range for this stretch while the output range will always be 0-255 (8bit). It's ultimately calling this function in rio-tiler (note the call to numpy.clip).

My data set ranges from 4.9m/s - 16.4m/s , when I use rescale am i correct in assuming that the data range min - max is split between 0-255 ? i.e. 0 represents 4.9m/s and 255 represents 16.4m/s ?

This is correct if you don't specify the rescale query parameter, in which case titiler falls back to scaling your data based on it's min/max range. If you wanted to cut off the histogram at 10m/s to remove outliers you could set the rescale query parameter to 4.9,10.

[mdavid800]

Thanks @geospatial-jeff that helps me understand it.

I assume if I set the rescale to 4.9 -10 m/s , everything above 10m/s would be set to 10m/s .

If I wanted to keep the outliers but just have the values > 10m/s coloured the same could I do that in the colour map?

[geospatial-jeff]

Yes I believe you would have to define a custom colormap that classifies values above 10 m/s with the same color. This would preserve the original values instead of clipping the image to the min/max values passed to rescale. You can pass a custom JSON encoded colormap to the colormap query parameter. Personally I've never done this before, so I'm not too familiar with the syntax.

[vincentsarago]

thanks @geospatial-jeff 🙏

@mdavid80, as Jeff mentioned you can either use rescale to rescale your input data from {min}/{max} to 0->255 and then apply a linear colormap OR you can directly use interval colormap https://github.com/developmentseed/titiler/blob/master/src/titiler/core/tests/test_factories.py#L136-L148