Jump to content


Photo

First Pixinsight Tutorial - Basic Image Processing with a little pixel maths!

Pixinsight tutorial images processing pixel maths video

  • Please log in to reply
34 replies to this topic

#1 Neil

Neil

    Administrator

  • Administrators
  • 1,493 posts
  • LocationThe Netherlands

Posted 14 December 2012 - 11:00 AM

This is our first video tutorial in the series on Pixinsight.

 

Pixinsight is a powerful astrophotography tool aimed at astronomers and the unique challenges that we face when processing our image data. In this first video we are going to process the Cocoon nebula. The data has already been calibrated and we focus on the next steps in producing an acceptable colour image.

 

 

We certainly hope that you find the video useful and informative.

 

Feel free to let us know what challenges or issues you face when processing your own images and we will try to cover them in future tutorials.

 

Clear Skies

 

 

Neil. 



#2 MickD

MickD

    Newbie

  • Members
  • Pip
  • 44 posts
  • LocationForest of Dean

Posted 14 December 2012 - 06:34 PM

Good tutorial - I see I'm going to have to join the dark side and get pixinsight, damm! more expense. Mick
  • JimMinCT likes this
WO FLT 98
Atik 383L
EQ 6 Pro
All in a RoR Shed thing.

#3 SlipperySquid

SlipperySquid

    Urban Legend

  • Administrators
  • 246 posts
  • LocationUnited Kingdom

Posted 15 December 2012 - 09:58 AM

Great tutorial matey, need to do morphological transformation next :D



#4 jlake

jlake

    Advanced Member

  • Moderators
  • 106 posts
  • LocationPomfret, CT

Posted 15 December 2012 - 02:12 PM

Great work, Neil! I had a couple of notes, and if you like the additions, perhaps they could be added as annotations in YouTube.

 

Notes:

  • An auto screen-stretch can also be done with Cmd-A (Ctrl-A on PC). I find myself going to this keyboard shortcut often during the linear phase of the image. For RGB images, it's best if the color channels get unlinked first (top left of STF process window)
  • I used to stretch manually (by eye) with the Histogram Transformation as shown in the video, but now I just apply the auto STF function to the histogram (by dragging the triangle from STF to the bottom bar of HT), then applying that directly to the image. This can cause for some seemingly-overstretched results at first, but it allows you to do a few more steps with the full range of data -- HDR, for one. Also, it allows the RGB and the L channels to be automatically balanced for the LRGB combination. Once I have that master, I do one more stretch to set the 'real' black point and mid. Here's Vicent Peris' (via a female American tech reader!) demonstration of the process. I think making the RGB and L masters compatible is a key step to good combination.
  • I also recommend combining the R, G, and B data first with the Channel Combine or LRGBCombination process (with L unchecked). Then you can work with the RGB master to extract the background and do color calibration, all while still linear. In fact, I think good color calibration MUST be done on linear RGB data.
  • Do you use the BatchPreprocessing script to generate your masters (or at least calibrate the data)? If not, I highly recommend it. One great thing it will do is Star Align all of the data to one single reference, so binned data will be scale up automagically! I think that running Star Alignment when making your master channel files, then running it again to get the masters matched, actually damages the data a bit. Every transformation will add some amount of error. Probably trivial amounts, honestly, but I've always tried to do one single transformation for any given file.
  • I was also a bit confused by the counterintuitive 'turning down' of the saturation slider in LRGB combination, but now I think it makes sense. It's the equivalent of a midtones slider, so moving it to the left curves it up, makes the saturation higher
  • You saved the Crop until the end, but if you were going to do any background extraction while still linear, all the edge artifacts would have to be cleared early. It's usually my first step once everything is registered.

Please take all this in the spirit that it's given: not in criticism, but in healthy difference that we'll all hopefully learn from.

 

If you would be able to make this data available (either here or through Dropbox, etc.), I would love to work with it. I could also make a screencast video of my different workflow, and perhaps by comparing and contrasting results, we could all benefit?

 

Anyway, as I said, great job on the tutorial. It's very clear and clean, and I think your steps are easily trackable. Thanks!



#5 Neil

Neil

    Administrator

  • Administrators
  • 1,493 posts
  • LocationThe Netherlands

Posted 15 December 2012 - 06:06 PM

If you would be able to make this data available (either here or through Dropbox, etc.), I would love to work with it. I could also make a screencast video of my different workflow, and perhaps by comparing and contrasting results, we could all benefit?

 

Anyway, as I said, great job on the tutorial. It's very clear and clean, and I think your steps are easily trackable. Thanks!

Not a problem Josh, your comments are really appreciated. I currently use MaxIm DL to capture and calibrate the data. I have used PI from time to time when MaxIm DL fails!!! (stacking only, not calibration) I'm still setting up and tearing down my kit each night, therefore, I'm alway introducing image rotation since I never get the ccd camera back into its previous position, henceMaxIm DL struggles. Let me know what data you want and we will find away to get it to you.

 

Personally, I'm alway intrigued to see what others can do with my data... This is after all the part of the hobby that is always changing and improving...

 

Regards

 

 

Neil.



#6 MickD

MickD

    Newbie

  • Members
  • Pip
  • 44 posts
  • LocationForest of Dean

Posted 15 December 2012 - 08:01 PM

Having been prompted by Neil's tutorial I have just spent some hours looking at Pixinsight on the net, very impressed....going to download trial version when back home tomorrow. Can see this going on my Xmas list. Mick
WO FLT 98
Atik 383L
EQ 6 Pro
All in a RoR Shed thing.

#7 jlake

jlake

    Advanced Member

  • Moderators
  • 106 posts
  • LocationPomfret, CT

Posted 16 December 2012 - 12:58 AM

Not a problem Josh, your comments are really appreciated. I currently use MaxIm DL to capture and calibrate the data. I have used PI from time to time when MaxIm DL fails!!! (stacking only, not calibration) I'm still setting up and tearing down my kit each night, therefore, I'm alway introducing image rotation since I never get the ccd camera back into its previous position, henceMaxIm DL struggles. Let me know what data you want and we will find away to get it to you.

 

Personally, I'm alway intrigued to see what others can do with my data... This is after all the part of the hobby that is always changing and improving...

 

Great, let's set up a Dropbox or some hosting. My Dropbox account is through my email, rootlake AT gmail. You could also host it on a site and email me the link.

 

As for what to process, I could go either way. if you want to give me the masters from Maxim, I'd be at the same exact starting point as the video, which could be good for comparison. If you wanted to host all the raw files in a big set, I could also demonstrate the excellent BatchPreProcessing script, which has made my life much easier!

 

Let me know, I'm looking forward to working with the data.



#8 Neil

Neil

    Administrator

  • Administrators
  • 1,493 posts
  • LocationThe Netherlands

Posted 16 December 2012 - 11:37 AM

Great, let's set up a Dropbox or some hosting. My Dropbox account is through my email, rootlake AT gmail. You could also host it on a site and email me the link.

 

As for what to process, I could go either way. if you want to give me the masters from Maxim, I'd be at the same exact starting point as the video, which could be good for comparison. If you wanted to host all the raw files in a big set, I could also demonstrate the excellent BatchPreProcessing script, which has made my life much easier!

 

Let me know, I'm looking forward to working with the data.

I have just setup a generic "Dropbox" and send you the link to the data... Can't wait to see the results...

 

I'll look into setup up a large file transfer system for members of the forum but it will probably take sometime... Something to do over Christmas.  :D



#9 jlake

jlake

    Advanced Member

  • Moderators
  • 106 posts
  • LocationPomfret, CT

Posted 17 December 2012 - 12:52 AM

Here's what I have so far on the Cocoon data. I'll admit that I did quite a bit more tinkering than I expected to tone down that 'salmon' color without wiping out the great detail from the H-alpha. Now that I know what I'm doing (sort of!), I'll try to record a video.

 

5KFBl.jpg?1



#10 jlake

jlake

    Advanced Member

  • Moderators
  • 106 posts
  • LocationPomfret, CT

Posted 18 December 2012 - 05:04 PM

Here's what I ended up with after some more noodling around. What do you think? Would it be worth making a video with some of the different steps I took?

 

7lBoa.jpg?1



#11 Neil

Neil

    Administrator

  • Administrators
  • 1,493 posts
  • LocationThe Netherlands

Posted 18 December 2012 - 10:00 PM

Thanks Josh,

 

Really, nice effort on the cocoon data, there is something to be said for both of them... I like the background in the first image and the details in the nebulosity of the second...

 

One thing that I realised after capturing the Cocoon is that I could have framed it better to show the inky dark lane leading away from the nebula, hence I do prefer the first image background because it shows the dark region better. The star colours are definitely better in the second image as well so a blend of the two of them would be perfect...

 

Really interested to see your video of the processing... I ask that I could have pushed it a little harder. In the old days of my image processing, like most I pushed substandard data too hard. Now I try to capture a lot of data and not push it so hard. However, I must admit that the STF (Screen Transfer Function) has done a good job here, if thats what you used?

 

 

Neil.



#12 jlake

jlake

    Advanced Member

  • Moderators
  • 106 posts
  • LocationPomfret, CT

Posted 19 December 2012 - 12:52 AM

It took a long time to upload, but here is the video: 

 

 

I feel like the background on the first one could be achieved by inverting my mask and using curves on the background -- a bit of de-saturation, a bit of contrast to dark.

 

Yes, I did use the auto STF stretch as a first step in order to make the H-alpha and RGB data match up. Once combined, I re-stretch the single HaRGB image to taste.

 

Let me know what you think about the process!



#13 Neil

Neil

    Administrator

  • Administrators
  • 1,493 posts
  • LocationThe Netherlands

Posted 19 December 2012 - 08:57 AM

Excellent work Josh,

 

And the Cocoon looks fantastic, even I'm learning from some of your steps. My overall comment is that you went a little too fast but people can replay and pause the video as needed.

 

What I liked and to add as a small aside to the video;

 

1, Short cuts for commands MAC = CMD 'A'; PC = CONT 'A' to apply a "Sreen Transfer Function"

 

2, "Space-bar" to scroll around the open windows to focus on the Cocoon, The "Window Tiling" to keep the desktop tidy.

 

3, RGB combine first and then apply the luminance data or H Alpha in this case. I would normally do this with galaxies but I think there is an underlying philosophy that we need to explore further...

 

4,Unlinking the RGB data to better see the actual background colour was an excellent step and even I would have potentially over looked it and probably used the "Back Ground Neutralisation Tool"

 

5, Copying the "Crop process" so that you can apply the same geometry to the HA. What we should tell everyone is that Pixinsight allows you to copy all of your processes / steps.

 

6, DBE = "Dynamic Background Extraction" Fully agree with your statement, this is also the one process that convinced me to hand over my money to the people and Pixinsight... Simply WOW... I like many in compromised locations get lots of gradients with my RGB data free of charge! In the past, it has cost me so much effort and time to find the third party tools to resolve this one problem that this alone makes PI worth its money.

 

7, STF = "Screen Transfer Functions" stretched are the same and this to is a time saver and something that I will be adding to my routine, so you have convinced me. I was a little sceptical because I thought its too aggressive but you have controlled it very well.

 

8, HDR Multi-scale Transform... Again Wow... This really brought out the dust in the Cocoon and in my opinion "transformed" the image...

 

9, SCNR wasn't really needed here but it can be used to remove a specific unwanted colour cast as you show. In most cases that colour will be "GREEN"

 

10, The "Range Selection" tool is also new to me and a nice way to generate a mask as you show. Most of the Tools in PI especially the  noise reduction tools come with their own ability to generate masks. I normally use the "Star Mask" tool for this purpose and adjust the scale etc... But it can be a bit hit and miss, so I will be using the "Range Selection" tool in the future.

 

11, And a "Morphological Transformation" to end it all... Brilliant.

 

Something for future video's should be the use of Previews in PI to test settinsg to optimise results and we should explain further the reasoning behind making the RGB image first and then adding the luminance data? I know that there is a valid reason but it escapes me at the moment.

 

Once again thank you very much for your efforts, it takes time to make these video's not to mention the courage! 

 

Josh and I will be announcing a PI challenge later, something to keep you occupied with over Christmas, especially if its cloudy.

 

Don't worry PI is free to try and you don't need to buy it if you don't already have it.

 

Neil.



#14 MickD

MickD

    Newbie

  • Members
  • Pip
  • 44 posts
  • LocationForest of Dean

Posted 19 December 2012 - 11:28 AM

Anyone got an idea how long it usually takes the "Pixinsight" guys to respond to a trial license request?

Sent it on Sunday, got the "we need to manually verify all requests" message, no further response, sent a gently reminder yesterday, still no response.

Am I expecting to much? It is Xmas.......

Mick
WO FLT 98
Atik 383L
EQ 6 Pro
All in a RoR Shed thing.

#15 Neil

Neil

    Administrator

  • Administrators
  • 1,493 posts
  • LocationThe Netherlands

Posted 19 December 2012 - 11:40 AM

Anyone got an idea how long it usually takes the "Pixinsight" guys to respond to a trial license request?

Sent it on Sunday, got the "we need to manually verify all requests" message, no further response, sent a gently reminder yesterday, still no response.

Am I expecting to much? It is Xmas.......

Mick

Normally, they respond in a day or two... so you should have it for Christmas... Mick.

 

But they are very busy at the moment with the next major PI development so it may bring an additional delay but I'd still expect an answer this week.

 

For those that don't know, PI will drop support for the 32 bit version focusing all their attention on the 64 bit version which is available for the PC, MAC and Linux.



#16 MickD

MickD

    Newbie

  • Members
  • Pip
  • 44 posts
  • LocationForest of Dean

Posted 19 December 2012 - 11:52 AM

Hope so, I finish work today (in about 10 mins in fact .....) got some data that I want to process but am holing out to use PI on it. Mick
WO FLT 98
Atik 383L
EQ 6 Pro
All in a RoR Shed thing.

#17 jlake

jlake

    Advanced Member

  • Moderators
  • 106 posts
  • LocationPomfret, CT

Posted 19 December 2012 - 01:59 PM

What I liked and to add as a small aside to the video;

 

Great tips and points all, Neil, thanks for the detailed viewing. Yes, I did move a bit fast -- it's tricky to pack it all in! Perhaps it would be useful to people to see just single tools explained slowly and carefully for future videos by us.

 

I think a processing challenge is a great idea and look forward to being involved!

 

Tell me a bit more about your views on combining L data with RGB? Do you always do the full LRGB in one go? I used to, but I was convinced of the separate processing pathway by Vicent Peris' YouTube video. He did a galaxy in that one, but I've found the technique to be incredibly effective on nebular processing as well. You basically get all the detail together in the L frame and do the color calibration steps in the RGB, then combine after stretching both to non-linear spaces. I'd like to hear your opinions on the philosophy of this vs. other methods.



#18 Neil

Neil

    Administrator

  • Administrators
  • 1,493 posts
  • LocationThe Netherlands

Posted 19 December 2012 - 02:57 PM

What I remember about L-RGB images is as follows;

 

1, Galaxies are very difficult to colour balance correctly and therefore, for these class of objects its best to separate the two processes as you have done.

2, In general with a 'L' luminance layer it is very easy to kill your colour data because of the integration differences, also you need to be careful with the colour data since it suffers from more noise normally.

3. We tend to shoot colour data 2x2 binned and luminance 1x1 or unbinned.

 

  • Now I think the problem is that if we combine the LRGB image at the beginning and find that it's colour is wasted out, when we increase the saturation we also increase the noise coming from the colour data and throughout the world image.
  • If we delay bringing these images together, creating the RGB first and then work on the 'L' and combine them at a later stage we have better control over the noise in the final image since the 'L' acts to smooth out the background etc...

The proof would be to perform both processes and to measure the amount of background residual noise in the image.



#19 Neil

Neil

    Administrator

  • Administrators
  • 1,493 posts
  • LocationThe Netherlands

Posted 19 December 2012 - 02:59 PM

Hope so, I finish work today (in about 10 mins in fact .....) got some data that I want to process but am holing out to use PI on it. Mick

Good man... Let me know and if it doesn't come I'll dig out my Windows license key for you. PM me for the details...



#20 jlake

jlake

    Advanced Member

  • Moderators
  • 106 posts
  • LocationPomfret, CT

Posted 19 December 2012 - 05:12 PM

Regarding binning at the exposure level -- I have read that it's a waste unless you have real sky-limiting issues, have super long download times, or need to save on hard drive space.

 

I've been shooting all my RGB unbinned at 1x1 for about a year now and have achieved good results. You only get the benefit of binning if you need the extra SNR to break out of thermal and read noise issues. I would imagine most of us don't have that problem.

 

Take a look at this discussion. The logic makes sense to me... 

 

Now software binning on the processing side, that's a whole other issue.







Also tagged with one or more of these keywords: Pixinsight, tutorial, images processing, pixel maths, video

Thank you for your interest in our Astronomy Forum...
It looks like you haven't joined us yet?
Why don't you register to get all the features of being a member! Its free!