Open lgbouma opened 2 years ago
Alternatively: for the smooth component consider linearly interpolating over a source-masked image, per Sec 3.2 of https://arxiv.org/abs/2111.15006.
Best of both worlds: look at goodness of fit metrics for all three approaches. (2dmedian+local annuli; tess-backdrop; linear interpolation).
Nb. these changes will need to be propagated to the BGV column in the fits files.
See: https://ssdatalab.github.io/tess-backdrop/
By eye, this looks like a better initial subtraction model than the 2d median filter that IIRC the pipeline is doing now.