Choose most appropriate interpolation method

NCAR / kcor-pipeline

Pipeline code for KCor

Other

3 stars 2 forks source link

Choose most appropriate interpolation method #172

Closed mgalloy closed 4 years ago

mgalloy commented 4 years ago

There is now an option to choose between bilinear and cubic for interpolations, realtime/cubic. Determine the best interpolation method to use for each interpolation done in the pipeline:

[x] applying distortion (cubic)
[x] fill gain, see below
[x] when using KCOR_FSHIFT (bilinear)
[x] when rotating by p-angle (bilinear)

To make the gain cause less of a cliff between inside/outside of the occulter, do:

find mean/median of annulus (over/under occult by 3) of gain
replace under occulter (over occult by 1) of gain with mean/median
use this as the gain to do the gain correction
try both cubic and bilinear interpolation for distortion correction

mgalloy commented 4 years ago

@jburkepile says:

The 20200417 data from the raw.fill directory are rotated slightly compared with the data on the archive from the production pipeline. The fill data are rotated a fraction of a degree in the clockwise direction.

mgalloy commented 4 years ago

For some unknown reason, the roll correction seems to 0.0 when using bilinear interpolation for the rotations using ROT. Why is this? Should the roll correction be changed to 0.0?

Can we use the Venus transit to pin down solar north?

mgalloy commented 4 years ago

@jburkepile is going to create a composite that will specify a new roll correction.

mgalloy commented 4 years ago

@jburkepile email:

Don made composite images of Kcor and AIA for Sept 10, 2017 with the sharp ray. The K-Cor images are processed with bilinear interpolation for rotation. It looks like 0.7 degrees (which is our current correction) is still (arguably) the best. I asked him to make very fine rotation changes, zero to 0.7 degrees in 0.1 degree increments. It is hard to see much difference between adjacent rolls but there is a clear difference between 0. and 0.7. I put the data on the shared MLSO folder:

Path: Mauna Loa Solar Observatory / K-Cor / KCOR roll correction / Bi-linear interpolation in processing /

I need to pick other days for composites to verify this. Mike ... I will ask you to process those days with bi-linear interpolation once I have a short list.