Workaround to overcome JPEG 2000 issue #139
Conversation
drnextgis
force-pushed
the
crop_mask
branch
3 times, most recently
from
f764fc0 to
4eba751
Compare
Codecov Report
@@ Coverage Diff @@
## master #139 +/- ##
==========================================
+ Coverage 89.41% 89.43% +0.01%
==========================================
Files 30 30
Lines 4279 4287 +8
==========================================
+ Hits 3826 3834 +8
Misses 453 453
Continue to review full report at Codecov.
|
drnextgis
changed the title
|
|
||
| # HACK: We have to overcome https://github.com/mapbox/rasterio/issues/1449 here | ||
| if self._filename.endswith('.jp2'): | ||
| masked = False |
There was a problem hiding this comment.
if mask is false, and in the orginal there existed a mask, we are not supporting this case?
There was a problem hiding this comment.
.jp2 doesn't support internal mask, so one of the possible way to mask all pixels that extend beyond the dataset's extent is overwriting masked to False: in this case crop returns np.core.ndarray and GeoRaster constructor converts it to masked array here. The drawback of this approach is that we masks legitimate 0 value pixels as well. The same issue you can observe while cropping GeoTIFF which doesn't contain any mask (internal, alpha or nodata).
|
This looks good to me: short and isolates the upstream issue. However, there's also the problem of the |
|
However, if I understand correctly your comment in rasterio/rasterio#1449 (comment), this isn't a JPEG2000-specific issue, is it? |
|
Yes, this isn't a JPEG2000-specific issue. Close this PR in favour of #141. |
Fix cropping for JPEG 2000 images.