evaluate the usefulness of Y band

pmelchior / des-exp-checker

Crowdsourced image quality control for the Dark Energy Survey

MIT License

9 stars 2 forks source link

evaluate the usefulness of Y band #55

Closed esheldon closed 10 years ago

esheldon commented 10 years ago

The new preliminary y1a1 image set includes Y band.

evaluate usefulness of the images
evaluate if the extreme difference of Y band properties compared to other bands will be confusing to users
evaluate highlighting Y band to the user so they are aware
evaluate filtering Y band from the exposures unless the user wants to see it

pmelchior commented 10 years ago

I've had a look at them, and I don't think they are too problematic. I suggest leaving them in without any further modification to the viewer.

pmelchior commented 10 years ago

OK, maybe my enthusiasm was premature since @kadrlica found plenty Y-band images and marked them with "Vertical star spikes". The good thing is that our new Hodge-podge page makes this very clear.

While I don't understand the cause of this, we should either add this to the tutorial or highlight the fact that these are Y-band images. Short poll on options 1. to 4.

I'm in favor of 3 (plus addition to tutorial)

kadrlica commented 10 years ago

If these spikes are not a problem (i.e., are uncorrectable) then they should probably be in the FAQ, not the tutorial. Unfortunately, the FAQ is a poorer place to advertise them.

esheldon commented 10 years ago

Can you post some examples?

I guess the choice is really between highlighting these so the user knows they are Y band and filtering them. I think highlighting is better.

kadrlica commented 10 years ago

I can write up something for the FAQ if @pmelchior agrees that this is the right place.

esheldon commented 10 years ago

I'm seeing a common trend in Y that the bright star masks are insufficient due to the large trails.

pmelchior commented 10 years ago

Yeah, the weird thing is that the trail is perpendicular to the readout direction. So, I suggest that we do three things:

Add a description to the FAQ for the star masking issue (to clarify that the mask only takes out the really problematic pixels (saturated or close to), but not the entire star. Special emphasis on Y-band there
Mention and link the FAQ in the relevant section of the tutorial (e.g. at Excessive mask)
I'll add the highlighting of the band when a Y-band images is loaded

esheldon commented 10 years ago

I remember Emmanuel had an explanation for this "trail". As I recall it is not actually a trail, but an optical effect that is in all bands but is most prominent in the Y band.

esheldon commented 10 years ago

(note they go out both "top" and "bottom")

kadrlica commented 10 years ago

Beyond the vertical star spikes, are there other features that need to be mentioned for Y-band?

pmelchior commented 10 years ago

They are more prone to fringing, but this should be taking into account by the flats, so if it doesn't work we actually want to know. Hence, no changes needed.

You can use the format we have in the tutorial with an image of the Y-band star masks, title something like "What's up with the stars in the Y-band". @esheldon Could you track explanation of Emmanuel down, please.

I suggest to clarify the situation of the size of the star masks in the tutorial, too: "How large should star masks be?"

kadrlica commented 10 years ago

I've started something in FAQ for Y-band.

I'm a bit surprised that we don't have a devoted problem type for star masks. Do we want to add one? This would be the place to address the star mask size question (I'm not sure I know the answer to this question).

pmelchior commented 10 years ago

I'm also not sure how the star mask sizes are determined. It's clear that they not only cover the saturated pixels and that they scale with magnitude, but often appear a little small. Let's take this discussion over to email.