Pixinsight - Colour Calibration problem

[Aramcheck]

I'm having problems in Pixinsight getting Color Correction on a recent image I took. I've tried both Photometric Color Calibration & standard CC, but my background keeps coming out almost white... Any ideas? I can only get something half-decent by using Manual White Balance, but I'm not too happy with the initial result.

The only difference I've made to my usual process is that I'm now guiding the scope & took 180sec subs.

Any help much appreciated!
Cheers
Ivor

[Anne S]

Have you tried applying stf again? That should adjust it.

[wimvb]

What are your background levels? You may need to increase the allowed level.

Alternatively, load up the tif here (non drizzled or it will be too large, or just a preview of the original) so we can have a look.

[Aramcheck]

1 hour ago, wimvb said:

What are your background levels? You may need to increase the allowed level.

Alternatively, load up the tif here (non drizzled or it will be too large, or just a preview of the original) so we can have a look.

RGB levels are 0.01 or below which is comparable to another image I took two days earlier, but with 60 sec rather than 180 sec subs. That earlier image went through PhotometricColorCalibation without a problem.

Not sure what settings to use to upload here as a tif, so here's a temporary link to the (non drizzled) linear XISF file with just DBE applied. It's 211Mb:-

https://drive.google.com/open?id=1BM0X42z7P_BNYP1scbiQotvZrqM2uRXn

Cheers
Ivor

[Aramcheck]

2 hours ago, Anne S said:

Have you tried applying stf again? That should adjust it.

Yes - first thing I did...

Cheers
Ivor

[wimvb]

58 minutes ago, Aramcheck said:

Not sure what settings to use to upload here as a tif

I just upload it as an ordinary image. Because no web browser supports tif or xisf, it will show as an icon.

[discardedastro]

Interesting - i think there's some issues with pattern noise which might be throwing off the automatic stuff. Just doing PCC (all default settings, using metadata from image) and then whacking colour sat up 3x to have a look at it gives the below:

You can also see lots of spread in the PCC charts - this isn't good, you should usually see a pretty tightly grouped set of points. There's not much in it in intensity between the fainter nebulosity and the background. How many frames stacked was this? Are you dithering the scope during guiding?

[wimvb]

15 minutes ago, discardedastro said:

How many frames stacked was this?

51 according to the fits header and the image identifier.

I found a fl of 1010 mm, not the 2000 of the original post.

[discardedastro]

6 minutes ago, wimvb said:

I found a fl of 1010 mm, not the 2000 of the original post.

Good spot - yes, I got that too. But doesn't appear to have a drastic difference on the processing.

[wimvb]

25 minutes ago, discardedastro said:

Good spot - yes, I got that too. But doesn't appear to have a drastic difference on the processing.

No, it doesn't.

This is what I got with default settings of PCC, with fl 1000 pixel size 4.3 and a background preview as background reference.

Arcsinh stretch and lifting the outer regions of the galaxy a bit. I see now that I boosted the colour in those regions a bit too much.

Processed in PixInsight 1.8.5

Btw, I would have framed the image a bit different to keep that small galaxy at the bottom in the frame.

Edited May 15, 2020 by wimvb

[Aramcheck]

1 hour ago, discardedastro said:

Interesting - i think there's some issues with pattern noise which might be throwing off the automatic stuff. Just doing PCC (all default settings, using metadata from image) and then whacking colour sat up 3x...

You can also see lots of spread in the PCC charts - this isn't good, you should usually see a pretty tightly grouped set of points. There's not much in it in intensity between the fainter nebulosity and the background. How many frames stacked was this? Are you dithering the scope during guiding?

@discardedastro Thanks - could you possibly share a screen grab of the PCC settings. I thought I had already used the defaults, but you've got a much better result!

It was 51 x 180 sec subs & 2nd night guiding. I haven't figured out what settings to use for dithering yet...

Cheers
Ivor

[Aramcheck]

1 hour ago, wimvb said:

I found a fl of 1010 mm, not the 2000 of the original post.

That's because the original screen grab is from a drizzled integration & doubling the focal length is necessary in order for the PCC to match stars. On the non drizzled stack I use 1000mm for focal length.

[Aramcheck]

37 minutes ago, wimvb said:

This is what I got with default settings of PCC, with fl 1000 pixel size 4.3 and a background preview as background reference.

Arcsinh stretch and lifting the outer regions of the galaxy a bit. I see now that I boosted the colour in those regions a bit too much.

Processed in PixInsight 1.8.5

Btw, I would have framed the image a bit different to keep that small galaxy at the bottom in the frame.

@wimvb Thanks - excellent, although I'm perplexed why my defaults are just giving a white image... Could you post a screen grab of the defaults?

In hindsight - yes, I should definitely have given the scope a nudge to bring that smaller galaxy into the frame. I had expected that I would need to crop the image so didn't bother.

Thanks!
Ivor

[Aramcheck]

@wimvb, @discardedastro - Thanks both. I've gone back to the non-drizzle stack, reset defaults and that now seems to be working ok...

I'll see how I get on now!

Many thanks

Ivor

[discardedastro]

You'll see pretty limited benefit, if any, from drizzling at 0.9px/" scale except with very good seeing.

I think you'll find neither me nor wimvb were posting screenshots with a STF applied - wim used a Arcsinh stretch rather than AutoSTF. You'll probably need to use a more careful stretch to draw out detail either in the final histogram etc stretch or in your working STF.

Also, do note that automatic STF can mess up white balance if not linked - worth checking to make sure you've got that set right.

Dithering, I'd just set the defaults in whatever you use and do a manual dither or two before you start to tweak. You want to achieve enough motion in each dither step that you're moving the objects at least 35+ pixels from their current position (for most apps which spiral or random-walk, meaning you can use pretty large motion steps and thus get a nice big distribution of sensor positions for any given astronomical object). If your step's motion is visible between two subs, you're good. That'll probably help a lot with rejection of more of the pattern noise, too, though you'll probably need to play with PI's integration rejection settings to get that gone.

[wimvb]

20 minutes ago, discardedastro said:

You want to achieve enough motion in each dither step that you're moving the objects at least 35+ pixels from their current position

For dslrs, 12 - 15 pixels is recommended (based on a talk held by Tony Halas). The problem with large steps is that it can start to affect stacking edges. If you do 50+ subs, those 35 random pixels can add to quite a lot, and you need to crop a wide edge after stacking. But as you say, you need to make sure that there is a clear difference between subs.

[discardedastro]

2 hours ago, wimvb said:

For dslrs, 12 - 15 pixels is recommended (based on a talk held by Tony Halas). The problem with large steps is that it can start to affect stacking edges. If you do 50+ subs, those 35 random pixels can add to quite a lot, and you need to crop a wide edge after stacking. But as you say, you need to make sure that there is a clear difference between subs.

Fair - I think it depends on the framing, too. I tend to be imaging targets small enough that the "border" I know I'll lose to dither overlap is pretty large with most of my images. Dither enough and capture enough, of course, and you'll end up with a wider usable field - sort of a small-scale mosaic!