Another M31. 8hr autosave added.

[MARS1960]

I'm rather disappointed with this, (i've seen Stuarts 2hr image), it took me three nights and a total of 8.3hrs captured, i stacked 73 of the best subs for a total integration time of 4.8hrs.

73 x 240s lights.

30 x flats.

30 x bias.

[gnomus]

M31 is often thought an easy object (I presume because it is so bright), but I think it is quite tricky (because it is so bright)!  Many folks will shoot a range of different length subs in order to try to prevent detail in the core from being blown out - possibly even as short as 30 seconds.  You then have to find a way of blending this core detail back in in a way that looks 'natural'.  It's interesting that you felt you needed to discard so many subs.  Why was that, I wonder?  Your stars look nice and round an dare not too large.  You have good colour in those stars.  You seem to have reasonable detail in the target - I am always lintrigued by those (near) parallel dust lanes at the edge of the galaxy (top right in your version).  

If you want a critique then .... There are some gradients which appear as patches around the galaxy - could these be improved with DBE or GradX or whatever you use?  (Was this shot during a time when Moony McMoonface was around?)  Colour balance is not quite there.  If you use PS's eyedropper (or similar in another program) it should give an indication as to where the imbalance is.  Finally, (and I think this is a significant issue) there is a strange effect in the brighter parts of the galaxy -  it almost seems that we are looking at something at the bottom of a swimming pool.  This is the lower left part of your M31:

I think this is caused by 'too much' of 'something', but I have no idea what - noise reduction maybe?  

I think it is certainly worth persisting with this data, because it seem sound, and I am sure there is more in there.  But I might go back a few steps before moving forward again.

Steve

[MARS1960]

Thanks for that Steve.

I discarded a lot of subs because FWHM was between 8-9, i only stacked the subs between 3-4.

I do have a tendency to overdo the noise reduction.

I've taken it back to PI and come up with this, still not good but better i think, i will however go back again and reprocess it with the tips you kindly gave me.

Mark.

 

[StargeezerTim]

I prefer the 1st version Mark. I use PS so can't really help with PI tips but in PS, I'd try building it up in small stretches, whilst protecting the overexposed bits with a mask or by fixing points on the curve. To me, the data looks fine, I think it needs a gentler touch processing and care maintaining the colour balance. Like you, I tend to go heavy handed on the smoothing, must get treated for that! 

[gnomus]

The colours in the first version are definitely better - this new one seems green and a bit anaemic.  Have you seen PI's new PCC tool yet? 

[MARS1960]

5 hours ago, StargeezerTim said:

I prefer the 1st version Mark. I use PS so can't really help with PI tips but in PS, I'd try building it up in small stretches, whilst protecting the overexposed bits with a mask or by fixing points on the curve. To me, the data looks fine, I think it needs a gentler touch processing and care maintaining the colour balance. Like you, I tend to go heavy handed on the smoothing, must get treated for that! 

I give up Tim. I spent over 10 hrs processing it 7 times in different ways, i used various masks with parameters to fit the actions and even followed Warren Kellers OSC workflow and still it looks nothing like Stuarts 2hrs worth . I don't get what i'm doing wrong as we both used a very similar setup .

5 hours ago, gnomus said:

The colours in the first version are definitely better - this new one seems green and a bit anaemic.  Have you seen PI's new PCC tool yet? 

I haven't Steve, now going to check it out, cheers.

[MARS1960]

This is the final process, slightly different workflow and much more subtle actions, i don't think i possess the processing skills to get it any better , but personally i think it's improved on the original.

[gnomus]

To my eyes, easily the best one yet.  

[MARS1960]

1 minute ago, gnomus said:

To my eyes, easily the best one yet.  

Thanks Steve, i'm still disappointed it looks nothing like what i wanted lol, but at least i've learn't a few things.

[wimvb]

23 hours ago, gnomus said:

  ...

Finally, (and I think this is a significant issue) there is a strange effect in the brighter parts of the galaxy -  it almost seems that we are looking at something at the bottom of a swimming pool.  This is the lower left part of your M31:

I think this is caused by 'too much' of 'something', but I have no idea what - noise reduction maybe?  

....

Steve

One way to get this effect in PixInsight is to overdo star reduction (morphological transform).

This process needs a VERY good star mask. Then in MT, use morphological selection, not erosion. Set to abt 0.25 in strength. Amount <0.5, iterations about 4 and a small element (5x5 + or x form). Do this a few times, changing the element from + to x and back.

[MARS1960]

1 hour ago, wimvb said:

One way to get this effect in PixInsight is to overdo star reduction (morphological transform).

This process needs a VERY good star mask. Then in MT, use morphological selection, not erosion. Set to abt 0.25 in strength. Amount <0.5, iterations about 4 and a small element (5x5 + or x form). Do this a few times, changing the element from + to x and back.

Thanks Wim, i'll have a go now and i'll drop midtones to stretch the star mask.

[gnomus]

I used to do that, but I am finding that a 'less contrasty' mask works well.

[StargeezerTim]

2 hours ago, MARS1960 said:

This is the final process, slightly different workflow and much more subtle actions, i don't think i possess the processing skills to get it any better , but personally i think it's improved on the original.

That is much better! I don't know how to do it in pi, but the colour balance needs tweaking to lower the magenta bias in the stars and lower the yellow in the galaxy and you have a fine image. You are right about processing, it is difficult. I have been doing this for three or four years and class myself as an improving beginner in processing. It does get easier with time but it is a skill that takes time to develop. The upside is that you will find yourself going back to earlier stacks and using new skills and approaches on them.

[MARS1960]

Thanks Tim, i'll do those tweeks you suggest and then save it to my galaxy folder.

[wimvb]

20 hours ago, MARS1960 said:

This is the final process, slightly different workflow and much more subtle actions, i don't think i possess the processing skills to get it any better , but personally i think it's improved on the original.

Why don't you post the process container for this image? That makes it much easier to see your workflow.

In PixInsight, show the process history of the image. Draw the small triangle onto the workspace. This creates the process container with all its settings for the image. Save the container and post (a link) here.

[MARS1960]

40 minutes ago, wimvb said:

Why don't you post the process container for this image? That makes it much easier to see your workflow.

In PixInsight, show the process history of the image. Draw the small triangle onto the workspace. This creates the process container with all its settings for the image. Save the container and post (a link) here.

Wim, it says process container empty.

All i did for that image was follow page 230 of Warrens book, but i skipped deconvolution and the second run of MLT, i finished with colour saturation.

I added in after TGVDenoise, range selection, unsharp mask and dark structure enhance.

[wimvb]

I'll have a look at that page.

Other methods to keep stars smaller, besides MT, are deconvolution, and masked stretch. In the latter, you need to keep the target background dark (less than default value), and increase rejection fraction. The default values will result in an image that is too flat.

If you just saved the image, and not the project, no process history will be saved when closing PixInsight. While you process, and if you save your work as a project, full history is available, and you can undo/redo an unlimited number of steps. Nowadays I always save my WIP as a project, just to have full history. I just uncheck 'save previews' when saving. I also try to delete unused masks and clones of images to save diskspace.

Btw, if you get odd coloured stars, the first suspects are usually colour calibration or background neutralisation.

[MARS1960]

Thanks for that Wim, just seen a tutorial on project saving so will do that in future and thanks for the masked stretch tip, thinking about it (i processed the image 7 times) i may have used masked stretch on the final image but didn't alter the defaults.

[gnomus]

It is very easy to end up getting 'locked' into a fixed way of doing things.  Not every image needs to be masked stretch.  Not every image needs NR!  Sometimes I would find myself doing a process because I 'knew' that I had to include this proces in my 'workflow' (what a word!!!), and the image would look worse.  I would then spend time with further processes trying to undo the damage that the first process had done.  I have tried to stop doing this.......

Masked stretch is sometimes useful, but sometimes leaves the image looking worse than a conventional stretch might do.  Try both and see what you prefer.  You could always blend the result.  You are using a DSLR.  If I use MS, I use it on the RGB image - I would hardly ever use it on Lum because, to my eye, it tends to lead to low contrast.  Take a critical look at whether or not a particualr process is really improving the image.  If not, don't use it....

In your latest version, the blacks are clipped.  In my opinion, there is also a purplish colour cast.  Saturation has been pushed too far.  The stars are much better than in V1, but they have a rather 'stark', 'crinkly' look to them that leads me to think they have been sharpened.  I would avoid sharpening stars (except in particualr circumstances).  

[MARS1960]

Thanks for that Steve.

I do tend to go through the same steps whatever image i'm processing and i must admit it's often rushed with little thought to the outcome.

I'm going to give M31 one last go using the whole 8.3hrs, take your comments on board and take my time, i used masked stretch only on V7 but this led to losing contrast which i think i tried to pull back a bit too much.

Cheers.

[StuartJPP]

1 hour ago, gnomus said:

It is very easy to end up getting 'locked' into a fixed way of doing things.  Not every image needs to be masked stretch.  Not every image needs NR!  Sometimes I would find myself doing a process because I 'knew' that I had to include this proces in my 'workflow' (what a word!!!), and the image would look worse.  I would then spend time with further processes trying to undo the damage that the first process had done.  I have tried to stop doing this.......

Masked stretch is sometimes useful, but sometimes leaves the image looking worse than a conventional stretch might do.  Try both and see what you prefer.  You could always blend the result.  You are using a DSLR.  If I use MS, I use it on the RGB image - I would hardly ever use it on Lum because, to my eye, it tends to lead to low contrast.  Take a critical look at whether or not a particualr process is really improving the image.  If not, don't use it....

In your latest version, the blacks are clipped.  In my opinion, there is also a purplish colour cast.  Saturation has been pushed too far.  The stars are much better than in V1, but they have a rather 'stark', 'crinkly' look to them that leads me to think they have been sharpened.  I would avoid sharpening stars (except in particualr circumstances).  

I can agree with this, since I only get to do 3 or 4 decent sessions a year (if that) my workflow is quite fluid (aka my memory is bad so I can't quite remember it). Sometimes I do a lot in PixInsight and then finish off in PS and other times (like my recent M31) I only used PI to integrate the subs, perform DBE and then an initial stretch...then it was off to PS.

The core of my M31 was burnt out slightly, but that is fairly easy to recover. I went through 10 iterations of processing the image, improving it each time slightly. Some tweaks are discarded as they add no value or make things worse, the benefit of PS and layers...easy to flip between new and old.

I follow some of the DSLR-LLRGB processing methods from Scott Rosen, though I only really follow a few things from that, mainly creating two files and processing them differently namely L & RGB.

I didn't use a masked stretch this time, I didn't feel it needed it, but I certainly can't remember all the processing steps I followed because I was initially only "playing" with the data until it started looking okay but by that time I had lost most of my history.

If you want to, you could post your stacked image and see if people want to have a go at it (or send it privately), with 8 hours worth of subs I'd expect there to be more hidden in it than is being revealed in the images you posted so far (though I could be wrong).

[MARS1960]

I was hoping you would post Stuart.

I doubt you are wrong going by your image Stuart, and we both used a DSLR.

Good idea on the stack, it will also give others a chance to practice some processing if they wish, so here is the 8.3hr stack if anyone wants to have a play.

 

 

Autosave.tif

[StuartJPP]

Mark,

Just a quick process.

Sequence was as follows:
PI
1) Dynamic crop
2) DBE
3) Histogram Transform
4) SCNR

Imported into PS and duplicate image and create and RGB and a Greyscale version (LUM)

Lum
1) Noise reduction (wasn't too bad)
2) Sharpen slightly
3) Select Lum and copy

RGB
1) Increase saturation in the RGB version (copy 3 layers, top layer Luminance, second layer Soft Light) x 3
2) Gaussian blur - quite agressive
3) Paste Lum above as new layer onto RGB image
4) Change blending mode to Luminance for that layer
5) Resized to 50%

That was it. I will have another go at it when I can.

[gnomus]

Very nice, Stu.   It's still upside down, though 

[StuartJPP]

6 minutes ago, gnomus said:

Very nice, Stu.   It's still upside down, though 

I agree it is upside down, I always post mine the right way up