M101 in PixInsight...views, opinions & advice?

[oymd]

Good evening everyone

With the weather being terible these past few days it was the perfect opportunity to do some reading of the Inside PixInsight book, in addition to some youtube tutorials.

I have been relying on helpful experienced forum members to process my data, so I thought I would go over some of my data myself, and have a go from start to finish with PixInsight.

Please advise if this final image looks ok?

I used:

Dynamic crop

Dynamic background extraction with 50 sample points

MLT

Co lour calibration and Histogram transformation

lastly color saturation

and TGVDenoise

 

Many thanks...

[tomato]

I am certainly no PI expert, I only purchased the license last week, but your workflow looks in line with the tutorials I have been following. You don't indicate if you have separate LRGB frames or if you have OSC colour data? I have the former, and apply DBE to each channel before combining.

Your final image is good but perhaps has a slight red bias, at least on my monitor? 

[MarkAR]

Looks pretty good to me, maybe a touch red.

26 minutes ago, tomato said:

I have the former, and apply DBE to each channel before combining.

That sounds like an idea I might try, just started with PI and saw there is a noise reduction script that is supposed to work well on gray scale images.

[RayD]

Very good effort.  Agree with Tomato on DBE to each of the LRGB stacks, and I would say all that your image appears to be missing, which you don't show in your workflow is some background neutralisation.  This will help make things a little more neutral, which should help with the slightly red cast.

You can also run SCNR on the red channel which may help.

Great start though.

Edited May 1, 2020 by RayD

[oymd]

50 minutes ago, tomato said:

I am certainly no PI expert, I only purchased the license last week, but your workflow looks in line with the tutorials I have been following. You don't indicate if you have separate LRGB frames or if you have OSC colour data? I have the former, and apply DBE to each channel before combining.

Your final image is good but perhaps has a slight red bias, at least on my monitor? 

 

22 minutes ago, MarkAR said:

Looks pretty good to me, maybe a touch red.

That sounds like an idea I might try, just started with PI and saw there is a noise reduction script that is supposed to work well on gray scale images.

 

8 minutes ago, RayD said:

Very good effort.  Agree with Tomato on DBE to each of the LRGB stacks, and I would say all that your image appears to be missing, which you don't show in your workflow is some background neutralisation.  This will help make things a little more neutral, which should help with the slightly red cast.

You can also run SCNR on the red channel which may help.

Great start though.

Thank you for your comments...

Of course..sorry, forgot to mention my acquisition data:

ED80 + .85reducer/flattener + OSC ASI 294MC Pro

60 subs @ 120s, 30 darks, 30 flats & 30 dark flats, stacked in DSS

Agree...not sure why its so red?

 

[MarkAR]

Maybe try PhotometricColourCalibration. 

[oymd]

1 hour ago, MarkAR said:

Maybe try PhotometricColourCalibration. 

just tried it...

Tried it like 10 times....Had no idea what I was doing. Got the data Starsbourg. Typed in M101, and clicked GET.

I just changed the FL to 510, and pixel size to 4.

It kept doing its thing every time, and eventhough it says at the end: successfully completed with no errors

Every time I get the message: Photometric calibration failed: Got 2 sample something, when FIVE are needed?

[MarkAR]

Probably not detecting enough of the right stars. I believe it works by plate solving.

[oymd]

2 hours ago, MarkAR said:

Probably not detecting enough of the right stars. I believe it works by plate solving.

Do you know how to get plate solve to work?

i simply cannot get it to work?

[MarkAR]

If you've tried all the different database servers and ticked Force Plate Solving then I'm not sure.

Maybe go back to when you Colour Calibrated and try it directly after that. 

If you've already tried that then I'm not sure why it wont work.

[oymd]

2 hours ago, MarkAR said:

If you've tried all the different database servers and ticked Force Plate Solving then I'm not sure.

Maybe go back to when you Colour Calibrated and try it directly after that. 

If you've already tried that then I'm not sure why it wont work.

Tried all the database servers...ticked FORCE PLATESOLVE, always comes up with the error at the end.

I took a screenshot 

 

[oymd]

Here is the log:

 

Starting Linear iteration.docx

Edited May 2, 2020 by oymd

[oymd]

8 hours ago, MarkAR said:

Probably not detecting enough of the right stars. I believe it works by plate solving.

It shows that is IS detecting all the stars it needs, but there is a FAIL at the end no matter what?

[oymd]

Ok, I think it finally worked.

I chose: DISTORTION CORRECTION, and unticked automatic limit magnitude, and chose 18 instead.

Was advised in Inside PixInsight book by Warren Keller.

Not sure of the result though?

What the >?)$!

 

[oymd]

Redid the whole thing...PCC seems to be working now. Used RangeSelection and made use of Masks...

It was interesting..to say the least...

 

[MarkAR]

Well done, colours look great.

There is so much to learn in PI but the results can be awesome, I did a process on my un-drizzled NGC3184 data last night starting off with Mure Denoise on the separate channels in grayscale. OMG what an amazing script. I then tried to get rid of some background gradients after RGB Combo using ABE and DBE and the results were horrendous.

Going to try the DBE before Mure to see what happens.

Edited May 3, 2020 by MarkAR

[wimvb]

On 02/05/2020 at 01:14, oymd said:

Agree...not sure why its so red?

Look at the histogram. I bet that the red peak is lower and wider than the green and blue. If so, you have more noise in the red channel. (If not, then I just lost that bet, and I'll buy you a pint whenever/if ever we meet.)

A wider histogram in red means that you have more bright red pixels in the background, even if all three peaks are at the same level. You also have more dark red pixels there, but those don't show (brighten) your image. Hence a red cast. Applying dbe or background neutralisation, generally don't remove this. Just create a mask that targets the background and apply scnr red or just decrease contrast in red with the colour saturation tool.

Btw, with an osc camera, I wouldn't bother splitting the channels and applying dbe on each. Dbe works just as fine on colour data as on mono. But what you can do, is apply mmt to chrominance just before you start stretching. Use 8 layers.

Edited May 3, 2020 by wimvb

[wimvb]

29 minutes ago, MarkAR said:

Going to try the DBE before Mure to see what happens.

I've so far never bothered with the Mure script, but afaIk, it must be used before any other processing in order for it to work. I know that Barry Wilson has a tutorial on this.

https://barrywilson.smugmug.com/PixInsight-Tutorials

 

Edited May 3, 2020 by wimvb

[oymd]

6 hours ago, wimvb said:

Look at the histogram. I bet that the red peak is lower and wider than the green and blue. If so, you have more noise in the red channel. (If not, then I just lost that bet, and I'll buy you a pint whenever/if ever we meet.)

A wider histogram in red means that you have more bright red pixels in the background, even if all three peaks are at the same level. You also have more dark red pixels there, but those don't show (brighten) your image. Hence a red cast. Applying dbe or background neutralisation, generally don't remove this. Just create a mask that targets the background and apply scnr red or just decrease contrast in red with the colour saturation tool.

Btw, with an osc camera, I wouldn't bother splitting the channels and applying dbe on each. Dbe works just as fine on colour data as on mono. But what you can do, is apply mmt to chrominance just before you start stretching. Use 8 layers.

Nice one!

Not sure how to look at the histogram, but I did invert the mask to protect the galaxy, and applied SCNR RED, and red is gone!!

Can you please clarify what you mean by: " apply MMT to chrominance just before stretching, use 8 layers" ?

Sorry, newbie here. Would appreciate if you can explain that in simple terms? Many thanks!

Here's the FIRST image with SCNR RED done. BIG difference. Thanks

:)

 

 

Edited May 3, 2020 by oymd

[MarkAR]

Looks much better.

Another tip I found for SNCR is to invert the image then apply SNCR then invert back, get rid of magenta.

[wimvb]

4 hours ago, oymd said:

Can you please clarify what you mean by: " apply MMT to chrominance just before stretching, use 8 layers" ?

http://wimvberlo.blogspot.com/search/label/MMT

Hope this helps