Option to apply vrt colormap to png

[cogeotiff]

Option to apply vrt colormap to png

[vincentsarago]

👋 Hi @kriscarle, thanks for opening this issue.

So to keep everything as simple as possible we could:

• add an option in array_to_img to apply a palette to a 1 band image (example).
• add a utils.get_palette function that accept either a rasterio dataset (as in geop) or a color_palette name (local palettes shipped within rio-tiler)

btw, there are some work to do to fully support one band output image (e.g. add mask) : https://github.com/mapbox/rio-tiler/blob/f33930f4da0493c6941e5c3e617364eb9caca86a/rio_tiler/utils.py#L354-L355

[kriscarle]

Hi @vincentsarago I was able to get it working like this, using rasterio's built in support for writing colormaps and MemoryFile. The PIL palette doesn't support RGBA values, plus I think this example can easily be adapted to support multiple bands.

import rasterio
from rio_tiler import aws
from rasterio.io import MemoryFile

def tile(event, context):
    args = event['pathParameters']
    x = int(args['x'])
    y = int(args['y'])
    z = int(args['z'])

    bucket = 'my-bucket'
    key = 'data.vrt'
    tile = aws.tile(bucket, key, x, y, z, 1, 256)
    with rasterio.open('s3://{}/{}'.format(bucket, key)) as src:
      colormap = src.colormap(1)
      meta = src.meta
      meta['driver'] = 'PNG'
      meta['height'] = 256
      meta['width'] = 256
  
      with MemoryFile() as memfile:
        with memfile.open(**meta) as dataset:
          dataset.write(tile, indexes=1)
          dataset.write_colormap(1, colormap)
        memfile.seek(0)
        return memfile.read()

[vincentsarago]

@kriscarle I'm not sure to follow 100%
I made this notebook to try to explain my though.

Basically your example does not imply any change in rio_tiler but it's more on the lambda handler side. What I think will be good is to add this in the documentation.

plus I think this example can easily be adapted to support multiple bands.

I'm not sure to understand the concept of applying color palette to multiple bands. I'll be interested to see any use case for that.

[kriscarle]

@vincentsarago

Yes, I ended up not needing to change rio-tiler, so it could just work as an example. Though it could still be a nice feature to automatically apply the color map from a VRT band if it is provided in the VRT.

I'm not entirely sure what I meant by multiple bands either. I think I meant that you can create tiles for multiple bands and then combine them in some way. Maybe for calculating NVDI on the fly?