Re: Implementing -scale functionality of gdal_translate


Sean Gillies
 

Hi,

On Wed, Feb 6, 2019 at 2:32 PM <linden.kyle@...> wrote:
So I'm seeing some odd behavior now. I ended up trying the masked_array solution, but I'm seeing breaks in my data at the tile boundaries.

Here's what the data looks like if I use reprojection with just the source band (code: https://pastebin.com/dZVRVKi2): https://imgur.com/WpFKfij

Here's what the data looks like I read into a masked array and use an ndarray source (code: https://pastebin.com/z1xd8mXu): https://imgur.com/2QkQQZf
Ironically enough I get the exact same result with my solution that doesn't use a masked array, but instead modifies the transform: https://pastebin.com/7ph1VYWX

Your code looks okay. I recommend padding your read window by a few pixels on each side so that tile_window, tile_data, and tile_transform are slight dilated. I think this is very likely to eliminate what looks like an off-by-one error in the reprojection, an error that you don't see when you use the entire source dataset as the reprojection source. Does that make sense?

--
Sean Gillies

Join main@rasterio.groups.io to automatically receive all group messages.