M45 - Struggling with Processing (well making up my mind)

[ianaiken]

Hi All,

Wonder if I can get some help me.  Firstly, maybe there isn't a problem, but I feel like there is.  I have processed my M45 data a lot now.  I think there is a problem with the calibration of the flats on the RGB data which is causing some issues with gradient around the outside of the image, anyway, the problem I have is related to how the final image comes out.  It always ends up looking kind of harsh, quite a bit of contrast and quite sharp.  I cannot really explain it better than this.  It just doesn't seem right (maybe it is I cannot decide).

If you could change something on this image what would it be, if anything?  I want to improve so any feedback is very welcome.  Do you think it's has too much contrast, too sharp, too unrealistic?  Maybe its the stars that make it look weird.  I know I should really try to fix the cores (once I figure out how to do that).

[ianaiken]

I guess what I see when I look at some other M45 images is that the dust looks like dust, but on this image it doesn't.

[Knight of Clear Skies]

I think it's a fine image with a lot of detail in the reflection nebulosity. I can't spot the gradient issues you mention but I'll have another look on my monitor at home. I can kind of see what you mean about the contrast and sharpness issues, it's like the bright streaks of nebulosity are sharper than the dust lanes whereas in other images it's the other way around.

I don't think the large stars are such an issue in images of Pleiades, after all they are very bright and fierce. It can be more distracting when there is an image of a nebula with one very bright star in frame.

There is a little bit of coma in the image, not sure if this could be reduced with a distortion tool.

Just my 2ps worth, hope that is of some help.

[JRWASTRO]

The image looks as if there is a strong colour cast / gradient that runs from L to R.

If you get rid of the colour gradient and rebalance the red channel because the gradient

removal seems to have washed out a bit of the red then you get this :

Jeremy.

[ianaiken]

Thanks Kinght.  Yes there is coma spacing is not right on the RGB.  I still need to get that sorted out.  As for the detail, maybe I need to mask differently (or just be different).  Let me know what you see when on a computer I would appreciate any further comments you may have.

Hi Jeremy - For the background, indeed there is a strong gradient still.  I think I missed some useful processing steps out.  I did DBE but it's probably not enough.  I ran Automatic Background Extraction in PixInsight just quickly there and this is what it came up with.

Resulting in

Think I need to step back a bit and do a bit more processing in the linear images.

[JRWASTRO]

Hello Ian,

I tried GradientXterminator and it did not work at all well here. So I created the colour gradient layer manually. Where I chose the colours and level.

I would like to know what criteria are used to distinguish between the colour cast and the required image.  I guess one would have to do a literature search.

Jeremy.

[lensman57]

Hi Ian,

Any Chance of dropping the unstretched data in DropBox and posting  a link please. I have both PI and StarTools and see what could be done. there is a lot of dust around M45 and is not at all blue so any DBE or ABE is going to get into trouble here.

Regards,

A.G

[ianaiken]

Definitely links below.  Yes there is a lot of brown dust too, which was on my original posting I think.  It's a tough one for sure.

Lum - https://dl.dropboxusercontent.com/u/1595099/SGL/M45/m45-lum-IanAiken.fit

RGB - https://dl.dropboxusercontent.com/u/1595099/SGL/M45/m45-rgb-IanAiken.fit

You will need to align the RGB to the Lum

[ianaiken]

Hi Jeremy - I would need to read up on the tool in the literature and forum to understand what it was doing.  It was the first time I used ABE in PI.  Usually I only use tools what I know what it's doing (at least as much as I can get my brain to understand).

[Knight of Clear Skies]

What a difference a monitor makes. Can see what you mean about the gradient issues now.

[ollypenrice]

Could you give us some clues as to how you processed this? PI or PS or a mixture?

I wouldn't believe your gradient map. It has far, far too much structure to be a gradient. It is reading the true structures in the image data and then treating them as gradient and mauling the data as a consequence. I'd suggest using an incredibly small number of markers in DBE. Just a handful where you have the best approximatin of background sky - which is a bit scarce in this target!

Olly

[lensman57]

Hi All,

Wonder if I can get some help me.  Firstly, maybe there isn't a problem, but I feel like there is.  I have processed my M45 data a lot now.  I think there is a problem with the calibration of the flats on the RGB data which is causing some issues with gradient around the outside of the image, anyway, the problem I have is related to how the final image comes out.  It always ends up looking kind of harsh, quite a bit of contrast and quite sharp.  I cannot really explain it better than this.  It just doesn't seem right (maybe it is I cannot decide).

If you could change something on this image what would it be, if anything?  I want to improve so any feedback is very welcome.  Do you think it's has too much contrast, too sharp, too unrealistic?  Maybe its the stars that make it look weird.  I know I should really try to fix the cores (once I figure out how to do that).

M45-LRGB.jpg

Hi Ian.

First of all may I say that you have the most brilliant  data for M45 or any other target that I have seen, it is just so damned clean. I just did a quick process of RGB data in PI, I did not use the L data and I just went for detail so the stars are not controlled. I think that there is something drastically wrong with your ABE or DBE as there is actually very little gradient to get rid of in your capture. I just used the DBE with some random samples in the corners and a few where I thought there was real background. I attach the outcome and I hope that u approve. So your data is spot on just take the processing easy.

Regards,

A.G

[lensman57]

Hi Ian,

Here is the version with the Lum added but for some reason the registration has failed, I really don't understand it . I will have another look at it tonight once I get some time. As I said it is very good data just need to take your time with it.

Regards,

A.G

[ianaiken]

Hi AG,

Yes it is good data.  Long and many subs help that.  Olly taught me that a couple of years ago on here.  It does mean you do not get through lots of targets, but that's OK they are not going anywhere and to be honest from my LP observatory I am happy to just focus on a few but do them justice as much as I can.  It's good to see your process is reasonably similar to mine.

Did you register the RGB against the LUM?  It should work fine in PI (if you are using that).  The Lum FOV is smaller than the RGB and I always do RGB matching Lum as not to change the Lum data in the translation process.

I processed both the Lum and RGB separately.  I think I used DBE incorrectly.  I used to just pick sample points that were completely avoiding nebulosity but recently I forgot that important factor (from having a break over the summer) and just let the tool place them for me.  They were quite spaced out too.  I will revist this.

My workflow is basically this with some variance depending on data:

1. Calibrate and Stack.

2. Dynamic Crop

3. DBE

4. ATWT for noise reduction using clone mask

5. Deconvolution on Lum (and sometimes on RGB)

6. Colour calibration on RGB

7. Masked stretch script merged with "STF stretch+HDR" and using techniques documented in http://pixinsight.com/tutorials/NGC7023-HDR/index.html along with some gradient compression (more so to try to see what it does and effects).  Good for controlling star size although I need to practise this process a bit more.

8. LHE

9. Curves

10 Hist

11. TGV

12. SCNR on RGB

Then lazily I used the LRVB-AIP script to combine L with RGB.  There may be better ways to do that.  I think I also used Morphological Transformation to reduce the stars a tiny bit.

I've recently got a crib sheet which contains a good consolidation of processing techniques from the PI forum and Jim's personal experience.  I am looking into this more.

My PI skill is still developing (I guess it always is) so I am working on understanding these processes more and more as I use them.  Another thing I struggle with is star cores and I know I need to do a smaller stretch and merge that to try to fix the cores.

[lensman57]

Hi AG,

Yes it is good data.  Long and many subs help that.  Olly taught me that a couple of years ago on here.  It does mean you do not get through lots of targets, but that's OK they are not going anywhere and to be honest from my LP observatory I am happy to just focus on a few but do them justice as much as I can.  It's good to see your process is reasonably similar to mine.

Did you register the RGB against the LUM?  It should work fine in PI (if you are using that).  The Lum FOV is smaller than the RGB and I always do RGB matching Lum as not to change the Lum data in the translation process.

I processed both the Lum and RGB separately.  I think I used DBE incorrectly.  I used to just pick sample points that were completely avoiding nebulosity but recently I forgot that important factor (from having a break over the summer) and just let the tool place them for me.  They were quite spaced out too.  I will revist this.

My workflow is basically this with some variance depending on data:

1. Calibrate and Stack.

2. Dynamic Crop

3. DBE

4. ATWT for noise reduction using clone mask

5. Deconvolution on Lum (and sometimes on RGB)

6. Colour calibration on RGB

7. Masked stretch script merged with "STF stretch+HDR" and using techniques documented in http://pixinsight.com/tutorials/NGC7023-HDR/index.html along with some gradient compression (more so to try to see what it does and effects).  Good for controlling star size although I need to practise this process a bit more.

8. LHE

9. Curves

10 Hist

11. TGV

12. SCNR on RGB

Then lazily I used the LRVB-AIP script to combine L with RGB.  There may be better ways to do that.  I think I also used Morphological Transformation to reduce the stars a tiny bit.

I've recently got a crib sheet which contains a good consolidation of processing techniques from the PI forum and Jim's personal experience.  I am looking into this more.

My PI skill is still developing (I guess it always is) so I am working on understanding these processes more and more as I use them.  Another thing I struggle with is star cores and I know I need to do a smaller stretch and merge that to try to fix the cores.

Hi Ian,

Yes I used PI and  and the work flow is more or less the same as yours . I just don't know why it did not register the Lum and RGB properly. RecentIy I  have had a cou[le of occassions that Star Alignment has failed to do the job properly. I extracted the L channel from the RGB in Lab mode and then recombined the Lum as L with a and b.

Regards,

A.G

[andrewluck]

Just used the colour data and this is my result:

1. Clone the image
2. Neutralise the background of the clone to make placing the DBE points easier
3. DBE on the original using the control points from step 2
4. Masked stretch
5. SCNR

I've been as minimal as I can be to avoid introducing processing artifacts. While there is some evidence of the dust around the main nebula I think you had a bad case of background mottling in your original post.

An aggressive stretch shows what looks like vignetting. Did you have an accurate master flat?

There's still some residual gradient. My feeling is that you need to look closely at your master flat but if that isn't possible then another iteration of DBE may help.

Nice contrast in the smaller stars even without a colour saturation boost though.

Hope this is helpful

Andrew

[ianaiken]

Thanks Andrew I will take a look at the DBE technique you used. The background dust is much more visible in the lum than the rgb. The gradient and master flat issue is something I did notice in the rgb. I cannot redo the flats but I can check out the files and maybe integrate them again etc...

[ianaiken]

The mottling may be from aggressive lhe

[ianaiken]

My DBE on the Lum is shoddy too.  OK so back to drawingboard on this to do the DBE correctly.  Thanks everyone for the help and guidance appreciate it!

[andrewluck]

As your main problem is vignetting you could try using Division as the correction method with DBE. This may work better.

Andrew

[ollypenrice]

Great to see these faint dusty reds coming through. Clearly this is a great dataset.

Olly

[ianaiken]

Thanks Andrew I'll have a look to see if I can correct for the poor flat.  It's strange looking back before the summer break I was using DBE correctly and manually assigning sample points.  Not sure what has got into me (other than watching more tutorials and probably getting confused over best practises) so that I was just using the sample generator and not correcting for bright areas or nebulosity.  I did some more processing on some other images I have processed just recently and the outcome is much better now I am using DBE correctly so I am very pleased I started this thread!

[ianaiken]

@Olly - well I don't think I am finished on it yet I may get some more data on this target, and some hopefully from a dark site rather than my LP home.  Since collecting this data I have corrected the tilt in the optics near enough and revised spacing so both sets of data should be better towards the edges.  I have 148 subs for lum some only 5 mins some 15.  I think I need to see what 30 min subs produces on both the Lum and RGB (from OSC).  I know the 30 min subs on the California Neb I am working on better than the shorter ones, the noise being lower for one thing.

Correcting my DBE approach I can see the difference below.  Exposure is 3 hours per pane and needs more.

Incorrect DBE sample placement

Quick but correct DBE sample placement on Pane 1.  You can see the background is much cleaner and accurate surrounding the target.  I could have pushed the large scale structures more but I just wanted to check DBE here.

Hope this thread is useful for others as it has been to me!

[lensman57]

My DBE on the Lum is shoddy too.  OK so back to drawingboard on this to do the DBE correctly.  Thanks everyone for the help and guidance appreciate it!

Hi Ian,

I did a bit more work on the data. So here it goes.I also attach a small crop of the capture as there is a little galazy out there that I have not seen before.

Regards,

A.G

[ianaiken]

Yes that galaxy is 550 million ly away a ugc something.