I have modified the code used in drp_stella in order to make a global fit to the continuum. I describe the steps below:

1. I manually extract the flux with purest tophat-extraction from the image. Example of the extracted boxes is shown in the extraction_focus_defocus.png image.

2. Then I fit the continuum with the Akima spline, drawing heavily from https://github.com/Subaru-PFS/drp_stella/blob/master/python/pfs/drp/stella/fitContinuum.py. The example of the results for one defocused image and focused image are shown in Continuum_DeFocus.png and Continuum_Focus.png images. Red shows the excluded data, black is the data that is fitted, blue shows the spline fit and yelow shows the residuals.

3. I use that function to remove the continuum from images. The example is shown in Removal_of_continuum.png, where I shown removal in continuum for a single spot at various levels of defocus. For the moment I have deemed the quality to be ``sufficient'', but this will soon be tested in realistic applications.

I believe that the quality of the subtraction is ``good enough''. This is based on the insight that I see no obvious influences of the continuum on the 2d modeling results of the F/2.8 and F/2.5 data taken in April and May of 2019. If the subtraction has to be improved further I will open new tickets.

Kiyoto Yabe see this ticket, it might be helpful