This page archives Slack comments from the splinter session on imaging background matching of the Improving JWST Data Products Workshop (IJDPW).
Macarena Garcia Marin
Hello @Thomas Williams. As discussed, for tomorrow's background imaging exercise the proposal is to test the background matching from your pipeline with a different dataset, and compare the outcome with the MAST data on the archive.
The proposed dataset is the Carina nebula, there is NIRCam and MIRI data. The MIRCam mosaic is larger.
Thomas Williams
very pretty! This is F335/F444/F770/F1130/F1280/F1800 with our background matching method, haven’t run the short nircams just to save time
Macarena Garcia Marin
Very nice Thomas! How does this compare with the MAST LVL3 data?
Thomas Williams
Not sure ! I will download ASAP
Macarena Garcia Marin
Many thanks to those who participated on the session.
We had a very productive imaging sky matching discussion, that often went into alignment but we will focus on that one tomorrow. Thomas presented very good Carina nebula results from his pipeline, and that was the perfect setup for the discussion. In summary:
- Thomas’ code does sky matching by using the “median of differences” as opposed to the “differences of median” the pipeline does. This seems to be doing a good job and would be worth testing it in the pipeline. All the pieces are there.
- Pipeline improvement proposal: The pipeline does skymatching and outlier as different steps; skymatching only uses a portion of the images. One suggestion to explore on the pipeline would be to:
- Get away with the skymatch step
- Combine skymatch with outlier detection. Outlier has all the information needed to execute skymatch using the entire imaging.
- Feasibility has to be evaluated by the pipeline developers (Mihai Cara) but we think it can help speed up the pipeline.
Varun Bajaj
As it turns out, Tom and I had the same idea of doing "median of differences"- I've implemented it here as a replacement to the default pipeline skymatch step. https://github.com/Vb2341/jwst-mosaic-skymatch
it works really well for MIRI, and uses the pipeline itself to do the reprojection, so it should be easier to implement etc if someone is looking to use the same algorithm
Here are the images that come out for F770W, F1800W, F1280W, and F1130W (I didn't yet align these perfectly, just well enough to make a clean enough level 3 image)
Note that if you dont compute the sky well i.e. using this step, a lot of the MIRI images end up getting 10-50% of the pixels rejected in the outlier step (throwing away a ton of the SNR you would get!). There are also often discontinuities in the sky
Macarena Garcia Marin
Great Results
