Faster Map Reprojection

[Alan Wang]

Hi all,

I'm currently trying to generate CEA projections of HMI magnetograms, but using map.reproject_to is taking around 90 seconds per image at a high resolution. Considering that I will need to reproject 50 maps at a time, it will take over an hour. Is there a way to optimize reprojection of sunpy maps?

Thanks in advance.

https://jsoc1.stanford.edu/SUM8/D1541341092/S00000/hmi.M_45s.20220819_235315_TAI.2.magnetogram.fits 

[Stuart Mumford]

Hello Alan,

reproject fundamentally isn't that fast, it's doing a lot of coordinate transforms and interpolation. Also nobody has spent any significant time profiling it and improving it's performance.

That being said, I think that your main issue here is that you are reprojecting to a very large array, and one which (I think) is oversampling your input image by a significant margin.

If you calculate the angular width of a disk centre pixel in heliographic carrington coordinates you get approximately 0.12 deg, in your example you are reprojecting to a carrington map with a pixel width of 0.02deg which is a significant oversampling. If you reduced the size of your array by about 3x you would get a significant speed improvement.

Stuart

--
You received this message because you are subscribed to the Google Groups "SunPy" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sunpy+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/sunpy/53c9fbdb-3eb2-47c5-b9a1-dad051316475n%40googlegroups.com.