On Friday, November 22, 2013 6:10:52 PM UTC-8, Stefan van der Walt wrote:
On Sat, Nov 23, 2013 at 4:00 AM, Michael Aye
If this is the driver that is exposed by GDAL, it's far too slow and crashy for HiRISE images.
These images look like they're about 1GB in size. Do you want to read the entire image into memory? How large would that NumPy array be? Are you interested in only loading a specific portion of the image?
I'm usually not interested in reading the whole image into memory, but would like to successively scan through the image in portions. But even that used to crash with the HiRISE images using open source JP2000 drivers, maybe because they use an exceptional harsh compression algorithm. Do I see it correctly, that the io.imread interface does not yet support partial reading of an image? Michael
On Tue, Nov 26, 2013 at 9:50 AM, Michael Aye
Do I see it correctly, that the io.imread interface does not yet support partial reading of an image?
afaik you are correct here. In fact, we've been wanting to refactor the io module dramatically for some time (search the list for discussions), but other aspects have taken priority. (To my chagrin yesterday, when I was having ludicrous amounts of imread-related errors. =\) Partial loads should also be considered in the refactor discussions. Juan.