Pelican Nebula - 8 hours H-alpha

[tooth_dr]

Here is my progress to date on the Pelican Nebula.  It is 8 hours of 600s subs.  I used APP to combine both DSLR subs and CCD subs in the same stacking process.

Any advice would be welcomed - does it look a bit flat or washed out?  The noise is evident around the periphery because the DSLR sensor is larger than the CCD, and it was slightly rotated.  I cropped off a minimum so some of the image is from less data.

 

Thanks for looking

Adam.

[symmetal]

Coming along nicely.  Not bad at all for not having "essential" cameras  You've just caught a mini-orion in the bottom left too.

Alan

[tooth_dr]

7 minutes ago, symmetal said:

Coming along nicely.  Not bad at all for not having "essential" cameras  You've just caught a mini-orion in the bottom left too.

Alan

LOL Alan, non-essential camera working ok indeed!  It's still a bit noisy I think, I wonder is it in the processing or the cameras themselves, or exposure times, or is it ok and I'm expecting too much?

[symmetal]

For comparison this is 4 hours of 300s subs on an ASI1600MM at unity gain. I just gave it a stretch in Photoshop. No other processing. In your image there are some black clipped areas which makes the noise around them stick out a bit more. You have 4 times the acquisition time of mine which other things being equal would imply a similar amount of data assuming you just used the red channel output. Mine's using a small aperture ZS-61.

Alan

[tooth_dr]

1 hour ago, symmetal said:

For comparison this is 4 hours of 300s subs on an ASI1600MM at unity gain. I just gave it a stretch in Photoshop. No other processing. In your image there are some black clipped areas which makes the noise around them stick out a bit more. You have 4 times the acquisition time of mine which other things being equal would imply a similar amount of data assuming you just used the red channel output. Mine's using a small aperture ZS-61.

Alan

Thanks Alan

My image is actually 4 hours of data from a cooled mono CCD, along with 4 hours from a DSLR, both with 7nm H-alpha filters.  It should be better.

[symmetal]

Ah, I thought this was a colour CCD. What does the CCD image look like without the combined DSLR. The edges of the picture do show the common DSLR blotchy noise and it's probably adding to the noise on the CCD image unless you've masked it off.

Alan

[gorann]

I would also expect that adding Ha data collected with an un-cooled DSLR (especially if it still have the Bayer matrix on) to cooled CCD Ha data would just deteriorate the image. Adding more data to an image only works if the new data is as good as the old, as far as I understand it. Maybe you want the DSLR data to increase the field of view but then it should probably only be used outside the areas covered by the CCD - but maybe that is what you have done.

[tooth_dr]

34 minutes ago, gorann said:

I would also expect that adding Ha data collected with an un-cooled DSLR (especially if it still have the Bayer matrix on) to cooled CCD Ha data would just deteriorate the image. Adding more data to an image only works if the new data is as good as the old, as far as I understand it. Maybe you want the DSLR data to increase the field of view but then it should probably only be used outside the areas covered by the CCD - but maybe that is what you have done.

Thanks for your reply.  I'm really an amateur, but to my untrained eye the combined image of both cameras looked smoother.  I will post all images tonight - DSLR only, CCD only, and DSLR and CCD both.   I used AstroPixel Processor and have the integration algorithm set to weighted, so that the better images contribute more to the final image.

[tooth_dr]

I dont want to use the DLSR to increase the FOV, it's not really much bigger than the CCD, it's only slightly longer.  The shorter length is practically the same.  I want to use it to contribute to the total data.

Unless I stack each separately and then combine the stack in PS, @Xiga mentioned this but I'm not sure how that would actually work?

 

[Xiga]

3 hours ago, tooth_dr said:

I dont want to use the DLSR to increase the FOV, it's not really much bigger than the CCD, it's only slightly longer.  The shorter length is practically the same.  I want to use it to contribute to the total data.

Unless I stack each separately and then combine the stack in PS, @Xiga mentioned this but I'm not sure how that would actually work?

 

Hi Adam

Gorann is right in saying that combining NB data from a DSLR would normally just deteriorate an image. But 4 hrs of data, even with a DSLR, is not to be sniffed at, and if done properly it should still be able to increase your overall SNR. 

Tell me, did you bung all your Ha subs (DSLR and 383L+) into APP and create one single Integration? And if so, what Debayer Algorithm did you use? (it is set in the first tab, number 0). You should be able to tell from the file name of the Integration. If it has 'Ha' in the title then you used the special DSLR Ha debayer algorithm. If it says 'AAD' then you used the normal debayer algorithm (for non-DSLR data or non-NB DSLR data). 

Now, if you used the AAD algorithm then your 383L+ subs were integrated correctly but your DSLR subs were not. And if you used the 'Ha' algorithm then your DSLR subs were integrated correctly but your 383L+ ones were not. 

The solution is to do 2 separate stacks, using the correct algorithm for each camera. If you don't, then you are definitely just adding more noise than signal to your final stack.  (tip - don't forget to use the Quality Integration setting for each stack).

Once you have your 2 stacks, you have a couple of options:

Option 1: APP. Try stacking the 2 stacks together. Experiment with the different 'Weights' settings in the Integrate tab to find which one works best. I would recommend testing the 'Quality', 'SNR', and 'Noise' ones. 

Option 2: PS. Apply an identical stretch to the 2 stacks and export them to PS. Layer the DSLR one on top, set the Blend mode to Screen and drop the Opacity down to 0%. You need to find the right % of the DSLR image that represents the tipping point whereby going above it only really adds more Noise then Signal. You can do this by drawing a rectangular marque box in a patch of background sky and zooming in. As you increase the Opacity, you will see the Mean (i.e Signal) going up in the Info panel, but you will also see the StDev (i.e Noise) going up too. Just determine at what point things are no longer improving. Don't be surprised if it's a very low Opacity amount that is needed to achieve this! When you think you've found the right amount, blink the DSLR layer on and off at various parts of the image (both high and low signal areas) at close range, and satisfy yourself that it is definitely an improvement. 

Hope that is of some help to you Adam. Let us know how you get on. 

 

[tooth_dr]

4 hours ago, Xiga said:

Hi Adam

Gorann is right in saying that combining NB data from a DSLR would normally just deteriorate an image. But 4 hrs of data, even with a DSLR, is not to be sniffed at, and if done properly it should still be able to increase your overall SNR. 

Tell me, did you bung all your Ha subs (DSLR and 383L+) into APP and create one single Integration? And if so, what Debayer Algorithm did you use? (it is set in the first tab, number 0). You should be able to tell from the file name of the Integration. If it has 'Ha' in the title then you used the special DSLR Ha debayer algorithm. If it says 'AAD' then you used the normal debayer algorithm (for non-DSLR data or non-NB DSLR data). 

Now, if you used the AAD algorithm then your 383L+ subs were integrated correctly but your DSLR subs were not. And if you used the 'Ha' algorithm then your DSLR subs were integrated correctly but your 383L+ ones were not. 

The solution is to do 2 separate stacks, using the correct algorithm for each camera. If you don't, then you are definitely just adding more noise than signal to your final stack.  (tip - don't forget to use the Quality Integration setting for each stack).

Once you have your 2 stacks, you have a couple of options:

Option 1: APP. Try stacking the 2 stacks together. Experiment with the different 'Weights' settings in the Integrate tab to find which one works best. I would recommend testing the 'Quality', 'SNR', and 'Noise' ones. 

Option 2: PS. Apply an identical stretch to the 2 stacks and export them to PS. Layer the DSLR one on top, set the Blend mode to Screen and drop the Opacity down to 0%. You need to find the right % of the DSLR image that represents the tipping point whereby going above it only really adds more Noise then Signal. You can do this by drawing a rectangular marque box in a patch of background sky and zooming in. As you increase the Opacity, you will see the Mean (i.e Signal) going up in the Info panel, but you will also see the StDev (i.e Noise) going up too. Just determine at what point things are no longer improving. Don't be surprised if it's a very low Opacity amount that is needed to achieve this! When you think you've found the right amount, blink the DSLR layer on and off at various parts of the image (both high and low signal areas) at close range, and satisfy yourself that it is definitely an improvement. 

Hope that is of some help to you Adam. Let us know how you get on. 

 

Brilliant Ciaran, thanks for this.  I'll try this all later.  I do feel that 4 hours of data from a DSLR with exif temp at 0-3oC should be useful, hopefully this is the case or you will see more of my DSLRs in the classifieds lol

[Xiga]

3 minutes ago, tooth_dr said:

Brilliant Ciaran, thanks for this.  I'll try this all later.  I do feel that 4 hours of data from a DSLR with exif temp at 0-3oC should be useful, hopefully this is the case or you will see more of my DSLRs in the classifieds lol

Yeah I'd be amazed if it doesn't improve things. Perhaps even by quite a bit. It's all just down to getting the stacks right first.

To give you some idea, for most of my NB images I usually only manage something in the region of 4 hrs of Ha myself. It can vary a little, but it's rarely significantly more than this. In fact, for my Pac-man Nebula picture (which i hope to post tonight) I didn't even manage 4 hrs of Ha, I think it was closer to 3 actually. Which probably goes a long way to describing why I had such a hard time processing it to my liking! lol

[don4l]

Have you processed both sets separately to do a comparison?

I think that Gorann's point about combining different data is correct.  I recently got a new Ha filter and adding in old data visibly degraded the final image. 

[tooth_dr]

57 minutes ago, don4l said:

Have you processed both sets separately to do a comparison?

I think that Gorann's point about combining different data is correct.  I recently got a new Ha filter and adding in old data visibly degraded the final image. 

I'm doing that now.  Both images are with a 7nm H-alpha filter, iirc your old filter produced large halos which isn't quite the same here.  I'm using two cameras with almost the same pixel size (5.4nm and 5.6nm) with the same scope and same filters.  The answer I guess will lie in what the data produces.

[tooth_dr]

6 hours ago, Xiga said:

You can do this by drawing a rectangular marque box in a patch of background sky and zooming in. As you increase the Opacity, you will see the Mean (i.e Signal) going up in the Info panel, but you will also see the StDev (i.e Noise) going up too. Just determine at what point things are no longer improving. Don't be surprised if it's a very low Opacity amount that is needed to achieve this! When you think you've found the right amount, blink the DSLR layer on and off at various parts of the image (both high and low signal areas) at close range, and satisfy yourself that it is definitely an improvement. 

 

 

Hi Ciaran.  I'm at this stage now, it's hard to tell when it starts to get worse BUT I think I have it!

[don4l]

17 minutes ago, tooth_dr said:

I'm doing that now.  Both images are with a 7nm H-alpha filter, iirc your old filter produced large halos which isn't quite the same here.  I'm using two cameras with almost the same pixel size (5.4nm and 5.6nm) with the same scope and same filters.  The answer I guess will lie in what the data produces.

By "degraded" I meant that the contrast in the nebulosity decreased.  When I blended the Ha into the Red channel, the weaker bits almost disappeared - I don't really notice the halos any more!!

I've looked at some of my old subs of the Pelican, and your data is noisy by comparison.  Maybe you had a night of bad transparency???

[tooth_dr]

6 minutes ago, don4l said:

By "degraded" I meant that the contrast in the nebulosity decreased.  When I blended the Ha into the Red channel, the weaker bits almost disappeared - I don't really notice the halos any more!!

I've looked at some of my old subs of the Pelican, and your data is noisy by comparison.  Maybe you had a night of bad transparency???

It could be, I don't remember it being particularly bad though, but that's not to say it wasn't!  More likely it is my processing and/or my stacking technique.

I've used Ciarans technique here, and stacked the two data sets separately.  Then I've layered the DSLR data on top in PS, and used screen at 16% opacity to blend it onto the CCD data.  

Do you think it looks any better?

 

[don4l]

Yes!

Much better.  That is much more like I would expect.  I'll read Ciaran's method later. 

[tooth_dr]

Another version here done using Ciaran's method - #1, where I stacked the stacks.  I did all the various methods of stacking, and after analysing all in PS, the weights version has the highest mean and lowest sd.  Comparison with first image processing.

 

 

[Xiga]

Hi Adam

Yes the new versions are a vast improvement over the one in the OP. Nice one! 

Looking at your comparison shots, each one has had a different stretch, so it is hard to see exactly how much the improvement is. After all, the one in the OP has had by far the biggest stretch, so it will always look noisier if for no other reason than that alone. 

So to try and better compare 'Apples to Apples' i downloaded all 3 Jpg's. The one from the OP (which i will call the Single Stack Method), the PS Method one, and the APP method one. I equalised the histograms as well as i could. They're not completely identical, but they're pretty close. I also rotated and cropped so they can be more easily blinked. As expected, both the PS and APP method stacks show a massive improvement over the Single Stack Method. As for which is bests between the PS and APP Method Stacks, that's much harder to tell, as the small differences could easily be explained by the slightly different histograms. It's too close to call for me, so I'll let you and the rest decide which one you think is best. ?

Just one last thing - you say you used the 'Weights' setting in APP. I've attached a screenshot of APP below, as you can see, 'Weights' is the Name of the setting. Under it there are various options to choose. Which one did you use, and did you try different ones? 

Single Stack Method (from OP): 

 

PS Method:

 

APP Method (Stacking the Stacks):

 

[tooth_dr]

7 hours ago, Xiga said:

Hi Adam

Yes the new versions are a vast improvement over the one in the OP. Nice one! 

Looking at your comparison shots, each one has had a different stretch, so it is hard to see exactly how much the improvement is. After all, the one in the OP has had by far the biggest stretch, so it will always look noisier if for no other reason than that alone. 

So to try and better compare 'Apples to Apples' i downloaded all 3 Jpg's. The one from the OP (which i will call the Single Stack Method), the PS Method one, and the APP method one. I equalised the histograms as well as i could. They're not completely identical, but they're pretty close. I also rotated and cropped so they can be more easily blinked. As expected, both the PS and APP method stacks show a massive improvement over the Single Stack Method. As for which is bests between the PS and APP Method Stacks, that's much harder to tell, as the small differences could easily be explained by the slightly different histograms. It's too close to call for me, so I'll let you and the rest decide which one you think is best. ?

Just one last thing - you say you used the 'Weights' setting in APP. I've attached a screenshot of APP below, as you can see, 'Weights' is the Name of the setting. Under it there are various options to choose. Which one did you use, and did you try different ones? 

Single Stack Method (from OP): 

 

PS Method:

 

APP Method (Stacking the Stacks):

 

Wow thanks Ciaran! Thanks for all that work. Weights was a typo. I meant I used all the weights settings, and Quality gave the highest mean lowest sd.  In terms of stretch, the last two were stretched with DDP on same setting

[jjosefsen]

Nice looking - the DSLR data definately improved it in the end here, quite smooth!

I find it hard to distinguish between the PS version and the APP version, perhaps a litle more detail in the APP version.