Can You/Should You Combine L-eNhance Data with L-eXtreme Data - Forums

Topic starter

Jun 1, 2023

Hi all,

I am currently working on a project of Gum 15 and surrounds. I do have a conundrum though.

One night I shot with my L-eNhance filter, whereas tonight I am shooting with my L-eXtreme filter due to the nearly Full Moon.

I am wondering if it is possible to combine L-eNhance data with L-eXtreme data? I hope it is as I would really like to benefit from combining my exposures from the other night with tonight's data with my L-eXtreme data.

If it is possible, how would I best go about combining the data?

Zak

Well Written Engaging

andrea tasselli

Frequent contributor

Jun 1, 2023

Of course it is, I just did it. Just add them up, no rejection, no filters and make sure you use LinearFit (in PI) between the two so that background mean value is the same between the two beforehand.

Zak Jones

Topic starter

Jun 1, 2023

andrea tasselli:
Of course it is, I just did it. Just add them up, no rejection, no filters and make sure you use LinearFit (in PI) between the two so that background mean value is the same between the two beforehand.

Awesome thanks for confirming! Just what I wanted to hear.

In that case, would I use ImageIntegration and combine the two images that way with the methods you described, or would it be possible to combine the data in WBPP?

Either way, I am very excited to see what comes out of this!

Zak

andrea tasselli

Frequent contributor

Jun 1, 2023

You need two processing pipelines as far as I have tried, that is you need to process the two data streams separately and add the integrated lights using PixelMath. It might be possible to process them in one go in WBPP but I don't use it (I do all the calibration "by hand") so I can't advise. If it is, you would still need to use LocalNormalization with one single frame reference for all of them at any rate.

Helpful Concise

Ben

Jun 1, 2023

Andrea, could you confirm (or correct) if I understood right that LN is to be applied to - in Zak's case - all L-eN subs using a L-eN reference, and same thing to all L-eX using a L-eX reference, and LF is to be used to match both integrated images prior to PixelMath combination and any stretching, BlurX, etc …?

Zak Jones

Topic starter

Jun 1, 2023

andrea tasselli:
You need two processing pipelines as far as I have tried, that is you need to process the two data streams separately and add the integrated lights using PixelMath. It might be possible to process them in one go in WBPP but I don't use it (I do all the calibration "by hand") so I can't advise. If it is, you would still need to use LocalNormalization with one single frame reference for all of them at any rate.

Thanks for your suggestions, I will try it and see how it goes.

I have been researching online and it is possible to use WBPP to combine two different images with different filters and stack them together, although I cannot try it until I am home this afternoon. I hope that the results are good!

The good news is that my total integration time is now just over 3hrs with my L-eNhance data and L-eXtreme data. Definitely need to add more integration time but it's a good start.

I did do a stack with just the L-eXtreme data and it turned out much better than I expected it to, plus minimal halos too so that's another bonus. I would like to edit it separately before attempting to combine my L-eNhance data with my L-eXtreme data.

I will report my findings once I have tried it all out.

Zak

Engaging Supportive

andrea tasselli

Frequent contributor

Jun 1, 2023

Ben:
Andrea, could you confirm (or correct) if I understood right that LN is to be applied to - in Zak's case - all L-eN subs using a L-eN reference, and same thing to all L-eX using a L-eX reference, and LF is to be used to match both integrated images prior to PixelMath combination and any stretching, BlurX, etc ...?

That's correct.

Zak Jones

Topic starter

Jun 2, 2023

andrea tasselli:
Ben:
Andrea, could you confirm (or correct) if I understood right that LN is to be applied to - in Zak's case - all L-eN subs using a L-eN reference, and same thing to all L-eX using a L-eX reference, and LF is to be used to match both integrated images prior to PixelMath combination and any stretching, BlurX, etc ...?

That's correct.

Ok thanks again for your input.

How would I combine them in PixelMath if I do end up using WBPP on each data set separately? For example, would I use the Average method or use Screen or something similar?

I have been reading online that I can do it with PixelMath or ImageIntegration by using the same images twice if it's under 3 images, like it is in my case.

Zak

Zak Jones

Topic starter

Jun 2, 2023

Zak Jones:
andrea tasselli:
Ben:
Andrea, could you confirm (or correct) if I understood right that LN is to be applied to - in Zak's case - all L-eN subs using a L-eN reference, and same thing to all L-eX using a L-eX reference, and LF is to be used to match both integrated images prior to PixelMath combination and any stretching, BlurX, etc ...?

That's correct.

Ok thanks again for your input.

How would I combine them in PixelMath if I do end up using WBPP on each data set separately? For example, would I use the Average method or use Screen or something similar?

I have been reading online that I can do it with PixelMath or ImageIntegration by using the same images twice if it's under 3 images, like it is in my case.

Zak

Forgot to ask which data set should I use as my reference for LinearFit?

My L-eXtreme data is much darker than the L-eNhance data. I tried to match the data when it came straight out of ImageIntegration without doing any DBE or cropping but it came out differently to what I was expecting.

Zak

andrea tasselli

Frequent contributor

Jun 2, 2023

Zak Jones:
How would I combine them in PixelMath if I do end up using WBPP on each data set separately? For example, would I use the Average method or use Screen or something similar?

I have been reading online that I can do it with PixelMath or ImageIntegration by using the same images twice if it's under 3 images, like it is in my case.

I would use Average, gives better SNR IMO. And, yes, you could do that trick but in that case use 4 files, the 2 images and their duplicates.

andrea tasselli

Frequent contributor

Jun 2, 2023 · Edited: Jun 2, 2023

Zak Jones:
Forgot to ask which data set should I use as my reference for LinearFit?

My L-eXtreme data is much darker than the L-eNhance data. I tried to match the data when it came straight out of ImageIntegration without doing any DBE or cropping but it came out differently to what I was expecting.

You should use the one with the smoother background, normally the L-EXT. But could be either. Most importantly it shouldn't have streaks/trails of any kind, nor haloes.

Zak Jones

Topic starter

Jun 2, 2023

andrea tasselli:
Zak Jones:
How would I combine them in PixelMath if I do end up using WBPP on each data set separately? For example, would I use the Average method or use Screen or something similar?

I have been reading online that I can do it with PixelMath or ImageIntegration by using the same images twice if it's under 3 images, like it is in my case.

I would use Average, gives better SNR IMO. And, yes, you could do that trick but in that case use 4 files, the 2 images and their duplicates.

Awesome thanks for your reply!

I am trying it with ImageIntegration but I get this error even though I have already duplicated both images:*** Error: Excluded 4 invalid file(s) - less than 3 images available for integration, aborting process.<* failed *>

I might not be doing it right, so I am not sure what I am doing wrong for this error to come up.

I also tried using PixelMath to combine the two images, but they were different orientations. I am going to try and use StarAlignment to try and align them before I try combining them again.

Also, would I use the images as they came out of ImageIntegration with no DBE or cropping, or would I do cropping, DBE then aligning them?

Thanks again for your insights, really appreciate it!

Zak

Zak Jones

Topic starter

Jun 2, 2023

andrea tasselli:
Zak Jones:
Forgot to ask which data set should I use as my reference for LinearFit?

My L-eXtreme data is much darker than the L-eNhance data. I tried to match the data when it came straight out of ImageIntegration without doing any DBE or cropping but it came out differently to what I was expecting.

You should use the one with the smoother background, normally the L-EXT. But could be either. Most importantly it shouldn't have streaks/trails of any kind, nor haloes.

Awesome thanks for this, will try and use the L-eXtreme as reference.

Zak

Zak Jones

Topic starter

Jun 2, 2023

So I finally got it to work!

I did the following:

1. StarAlignment on both subs, with my L-eNhance data serving as reference.
2. LinearFit on my L-eNhance data using my L-eXtreme data as reference.
3. Combined the two images using PixelMath with the following expression: Image1 + Image2 with the average combination method.

It turned out much better than I expected it to! The amount of details I have managed to bring out of the nebulae with both sets of data has blown my mind. I will definitely be doing this more often should I happen to use different filters on the same target over different nights.

I am also running another experiment with WBPP just to see if the results are similar or different. I have grouped the two data sets together using keywording to separate the images by night taken. Once WBPP is finished (including LocalNormalization), I will stack the images together in the morning and I will advise my results once that is finished. Might even include a screenshot of both images side-by-side for everyone's reference.

Zak

Helpful Engaging

andrea tasselli

Frequent contributor

Jun 2, 2023

Zak Jones:
I am trying it with ImageIntegration but I get this error even though I have already duplicated both images:*** Error: Excluded 4 invalid file(s) - less than 3 images available for integration, aborting process.<* failed *>

I might not be doing it right, so I am not sure what I am doing wrong for this error to come up.

I also tried using PixelMath to combine the two images, but they were different orientations. I am going to try and use StarAlignment to try and align them before I try combining them again.

Also, would I use the images as they came out of ImageIntegration with no DBE or cropping, or would I do cropping, DBE then aligning them?

You should integrate from a set of files, with different names obviously. The two indipendent L-ENH and L-EXT integrations and their dublicates, as separate files. The two images should be aligned with each other and once you add them up you would do the required cropping. I assume that each data stream would produce an integrated light which is as pristine as it can get, i.e. with the background flattened.

Tim Hawkes

Jun 4, 2023

It is quite often the case that one might wish to combine two or more integrations from different observation sessions together (be they each obtained with same filter or - as in your case - with slightly different filters).

Further to the above suggestions - using linear fit and dynamic alignment before adding them together in PixMath - which all looks very sensible to me it might also be an idea to think about the relative PSF signal to noise weightings (or SNRs) of the images? The default in the normal PI Integration process is to combine frames based on their PSF signal weightings. By analogy it might be worth pre-running subframe selector just to provide an idea of the relative PSF signal weights of the two integrations that you wish to add together in PixMath? Then - say that image 1 has a PSF weight 2.5 x higher than image 2 - you would use a Pixmath expression "2.5X image 1 + image 2" rather than just "image 1 + image 2" and then make sure that the "rescale" option is checked in PixMath.

Tim

Helpful Insightful Respectful Engaging

andrea tasselli

Frequent contributor

Jun 4, 2023

I would be cautious in using PSF weighting with two different (significantly) filters such as these, as the strength of the PSF signal isn't strictly correlated to that of the object you're imaging.

Well Written Concise

Tim Hawkes

Jun 4, 2023

andrea tasselli:
I would be cautious in using PSF weighting with two different (significantly) filters such as these, as the strength of the PSF signal isn't strictly correlated to that of the object you're imaging.

Good point Andrea. Maybe weighted average based on the SNR of a key feature of e.g. a target nebula - but that needs more measurement and tailoring.

Joe Santacroce

Jun 4, 2023 · Edited: Jun 4, 2023

I do it all the time in APP and PI in a single stream with no issues. In WBPP I set up group keywords for the session and/or filters. You can have it process as two separate masters or a single master.

In APP (Astro Pixel Processor) there are prompts for multi-session and/or Multi-Filter.

After Blinking through the images I usually pick an L-eXtreme image as the reference image.

You can of course process separately and combine using a percent if each that you want.

Mark Germani

Jun 7, 2023

I did this recently with L-eNhance and L-eXtreme data using APP as @Joedoes, and I was happy with theresult. I do it all the time with broadband targets with and without an L-Pro filter, so I figured I could just combine the L-eXtreme and L-eNhance into one stack. APP does some normalization during the stacking process, and whatever it did seemed to work just fine.

I'm about to try and combine new L-eXtreme data withAstronomik CLS data taken two years ago. It'll be very, very interesting to see if my old data contributes in a positive or negative way to the final result.

CS,
Mark

Helpful Engaging

Zak Jones

Topic starter

Jun 8, 2023

andrea tasselli:
Zak Jones:
I am trying it with ImageIntegration but I get this error even though I have already duplicated both images:*** Error: Excluded 4 invalid file(s) - less than 3 images available for integration, aborting process.<* failed *>

I might not be doing it right, so I am not sure what I am doing wrong for this error to come up.

I also tried using PixelMath to combine the two images, but they were different orientations. I am going to try and use StarAlignment to try and align them before I try combining them again.

Also, would I use the images as they came out of ImageIntegration with no DBE or cropping, or would I do cropping, DBE then aligning them?

You should integrate from a set of files, with different names obviously. The two indipendent L-ENH and L-EXT integrations and their dublicates, as separate files. The two images should be aligned with each other and once you add them up you would do the required cropping. I assume that each data stream would produce an integrated light which is as pristine as it can get, i.e. with the background flattened.

Good points, thanks for the feedback! I will try this method once I start collecting more data for my L-eXtreme subs.

What I ended up doing was stacking both datasets together in WBPP, which turned out great. I did try and use your method of using PixelMath and averaging the two images together, although I haven't edited it yet, which I will probably do over the weekend as it's cloudy on Saturday and Sunday where I live.

Zak

Zak Jones

Topic starter

Jun 8, 2023

Tim Hawkes:
It is quite often the case that one might wish to combine two or more integrations from different observation sessions together (be they each obtained with same filter or - as in your case - with slightly different filters).

Further to the above suggestions - using linear fit and dynamic alignment before adding them together in PixMath - which all looks very sensible to me it might also be an idea to think about the relative PSF signal to noise weightings (or SNRs) of the images? The default in the normal PI Integration process is to combine frames based on their PSF signal weightings. By analogy it might be worth pre-running subframe selector just to provide an idea of the relative PSF signal weights of the two integrations that you wish to add together in PixMath? Then - say that image 1 has a PSF weight 2.5 x higher than image 2 - you would use a Pixmath expression "2.5X image 1 + image 2" rather than just "image 1 + image 2" and then make sure that the "rescale" option is checked in PixMath.

Tim

Thanks for the feedback Tim, I will look into doing this when I can, most likely over the weekend.

For Andrea's method, I did use LinearFit to set the image brightness to match the L-eXtreme dataset, but that was after first using StarAlignment to match the alignment of the frames. I then averaged the two images in PixelMath and it produced an image which I haven't edited yet, but from the looks of things, it was great!

Zak

Helpful Respectful

Zak Jones

Topic starter

Jun 8, 2023

andrea tasselli:
I would be cautious in using PSF weighting with two different (significantly) filters such as these, as the strength of the PSF signal isn't strictly correlated to that of the object you're imaging.

Thanks once again for your inputs Andrea!

Will take this into consideration for future edits where I am in this type of situation again where I use two different filters.

Zak

Zak Jones

Topic starter

Jun 8, 2023

Joe Santacroce:
I do it all the time in APP and PI in a single stream with no issues. In WBPP I set up group keywords for the session and/or filters. You can have it process as two separate masters or a single master.

In APP (Astro Pixel Processor) there are prompts for multi-session and/or Multi-Filter.

After Blinking through the images I usually pick an L-eXtreme image as the reference image.

You can of course process separately and combine using a percent if each that you want.

I haven't used APP yet, although I might try it out one day as it's got some good tools in there that are good alternatives to PixInsight, plus it saves massively on storage space!

Zak

Zak Jones

Topic starter

Jun 8, 2023

Mark Germani:
I did this recently with L-eNhance and L-eXtreme data using APP as @Joedoes, and I was happy with theresult. I do it all the time with broadband targets with and without an L-Pro filter, so I figured I could just combine the L-eXtreme and L-eNhance into one stack. APP does some normalization during the stacking process, and whatever it did seemed to work just fine.

I'm about to try and combine new L-eXtreme data withAstronomik CLS data taken two years ago. It'll be very, very interesting to see if my old data contributes in a positive or negative way to the final result.

CS,
Mark

Good stuff! Glad to hear that your experiment went well with APP.

Please let us know how it goes trying to combine your L-eXtreme data with your Astronomik CLS data, as I'm keen to know how it turns out!

Zak

Well Written Respectful Engaging Supportive