Add sky map pad method #624
Conversation
| @@ -387,6 +387,48 @@ def cutout(self, position, size): | |||
| skymap = SkyMap(data=cutout.data, wcs=cutout.wcs, unit=self.unit) | |||
| return skymap | |||
|
|
|||
| def pad(self, factor, mode, **kwargs): | |||
There was a problem hiding this comment.
I think for a general public padding method this factor argument might be too special. It is hard to understand from the docs how it is actually used. Looking at the code I presume it something like:
Pad image to the nearest larger shape, that is divisible by the given factor in both axis.
For a general purpose padding function I find a pad_width argument much more intuitive.
|
@OlgaVorokh will continue this PR in #630. Closing this now. |
| (16, 12) | ||
| """ | ||
| pad_width = factor - (np.array(self.data.shape) % factor) | ||
| pad_width = [(0, pad_width[0]), (0, pad_width[1])] |
There was a problem hiding this comment.
Thinking a bit more about this, I still prefer symmetrical padding (i.e. on all edges), so that the center of the image is unchanged (or at least add this as an option). The down- and upsampling can create artefacts at the data edges, depending on the boundary handling, that is chosen. I would find it weird, if this only affected the right and upper edge of the image. I would rather expect a consistent treatment of all the edges, e.g. how it's done for convolution operations.
And typically users will have a source in the center of the map. If the map is padded (for whatever reasons...), this will move the source out of the center of the map. Which is not that nice e.g. when it's plotted.
This pull request adds a pad method to
SkyMap, to help implementSkyMap.downsamplein #615 .Coded with @OlgaVorokh