Can't tell you how much I appreciate your tutorials. I'm also fairly new to it and every time I get a little lost I go back to these as references. I don't seem to have the issue you're having between Graxpert and Siril though. Not sure why but mine looked the same in Siril as I left them in Graxpert. Thanks again. I don't subscribe to many channels, but you've earned this guy's click.
A great ‘refresher’ video for me Rich. I’ve been using Siril and Graxpert for a while now after watching your earlier tutorials and I endorse what you say early on about how the unstretched image looks very odd! I don’t know the answer. I’ve never got to terms with ‘Deconvolution’ and must now give that a try. Thanks also for the tips you include about GHS and how best to preserve colour. I’ve got to give that a try too. Thanks as always 🤝
Yeah that's probably an easier way to go about it. Too bad right now we can't denoise while still linear. Pretty sure they recommend denoising before stretching, but like you said, it's a pain.
I too didn't have this problem with the latest version of Graxpert. However, I could not plate solve the saved image in Siril, so I colour calibrated and removed green noise before using Graxpert. Not sure if that makes a difference. I only used 10 x 300s exposures on the lobster nebula, so there was a lot of noise. Graxpert did an excellent job removing this. Thanks for this. UPDATE: I got better results applying Graxpert to the starless image (without cropping), processing in Photoshop, then recombining in Siril.
@@DeepSpaceAstro I have used this workflow on three sets of data so far and it has been AMAZINGLY helpful. I have been able to stretch out so much detail I didn't even realize was there when I tried processing this stuff a handful of months ago. Thanks so much.
Funnily enough your workflow matches mine. That's because I didn't have one and now I do! Your channel is a Godsend, much appreciated and keep 'em coming!
Questions/comments: I've noticed the deconvolution problems but hadn't attributed them to GraXpert. With small Seestar images, I find cropping is not necessary. Process is fast. Thanks for tip on reducing no. of iterations. In adjusting step size, try initial step increase to .0004 or .0005 to save time. In another video, you said AstroSharp may conflict with deconvolution. Any update? Nice tip about color stretch model. Is that true for nebulae too? I notice you did not use symmetry pt. for first 2 stretches. In linear stretch, you can click on histogram to set black point. Towards the end of processing, I try to make black area ~7-10% saturation. I noticed you did not use Highlight protection for galaxy core & didn't check the saturation there. I've been trying to avoid overstretching of bright areas (and bright stars) when I stretch them. In final image, sometimes stars have a green cast (perhaps due to independant color stretch?), so I use remove green noise tool. Also, I often increase saturation and/or do a saturation stretch rather than using my old (CS4) version of Photoshop. Sometimes after Siril processing, I use another noise reduction in GraXpert and/or the last two noise reduction processes in Siril. But occasionally these introduce serious artifacts in a galaxy core. Perhaps from GraXpert?
It's not just AstroSharp, you probably shouldn't combine any sharpening tools together, but you can always try and see what happens. Regarding the rest of your comments, there's not one way to do anything. Experiment with the tools to get the results you're happy with.
Hello, I'm using Linux Version of both software. No strange stretch problem when I re-open fits in Siril. It seems to works great! After using GrapXpert I have 6420 stars in my image instead of 2300 (using only Siril)
Well, I know what I'll be watching over my lunch break. Siril is a powerful package, but it's so underused in my worflow since I only use it for stacking, background normalisation and removing green noise. I should really do a deep dive into the documentation one of these days...
@@DeepSpaceAstro Yup, I noticed you mentioned it in this video, so I'm already watching it. So far, it looks more or less like the process I'm already following, which I got from Nico over at Nebula Photo's. Except from what I gather, you're not leaving Siril to post-process in Gimp or Photoshop. So curious to see what that workflow looks like.
Im curious what you might do different for pure clusters, like opens. I’m shooting at 1950efl and I’ve gotten some good globular details but I haven’t really landed on a foolproof fundamental approach. Starnet+ is a core part of my Galaxy/nebula workflow, and in Recomposition my goal is to keep the star fields from overwhelming the subject. But for clusters, the stars are the show. Much tougher I find.
@@DeepSpaceAstro I love it. Very similar to my basic workflow but I note the differences, and they make good sense. I wonder if those 'galaxies' could be saved by doing a stretch, then taking that starless into PS as a tif and with Color Range Select Shadows getting a mask over the blown-out core and sliding the Exposure down, re-saving into Siril as a fit then using that for a Recomposition. I suppose that could create a nasty artifact the other way though?
Hey Rich, thanks for sharing. Not having read all the comments - have you tried using Graxpert from within Siril? I don’t get that “weird stretching” when I use it that way. Also I’m using the alpha-dev 1.3.2 version of Siril.
Hey, Rich, thanks for another on-point video tutorial. I have also been experiencing some WEIRD side effects when hopping back and forth between Siril and GraXpert. There seems to be some kind of mismatch between the two programs. Most recently, I noted it when processing a wide-angle image of the Veil Nebula complex with that dense star field surrounding it. I am not sure if the problem is directly related to the order of operations or what. I was trying to use the AI versions of GraXpert gradient removal and denoising. Weird.
Yeah it's the changes it makes to the linear data that makes the Autostretch too much afterwards. The data itself is ok. For now we just have to deal with it.
...After Graxpert I get back to Siril and stay only in the "Line- Mode", stretching my image there very lightly with the Asinh Transformation. An other good way is the Arcsinh- Curves- adjustment- workflow at Photoshop. After all I go ahead as I learn from you: PSF...Deconvolution... Starnet++ and so on. Greetings from Thailand
Great video my friend,I always use siril for astro and your tutorials are always very helpful. I've got a 130mm reflector I use for astro but have been unable to get a decent image of a globular, a tutorial on capture and processing in siril would be super useful if possible... many thanks
Hi Rich, thank you for your channel as it is really very helpful to me. But I have a little problem and I would like to make some points. The moment I recompose the starless and starmask files, and save in Tiff the final file, I notice some loss of detail and color. So I thought of this solution, which at least works for me. Considering that I don't trust the star recomposition process nor Siril's export process, I open the two Fit files with GraXpert, export them to Tiff and work them in Photoshop creating two layers and applying screens to the star layer and...boom...the image is now as perfect as I expect !
Hello Rich, your video is very good! It's more or less how I've been working, you clarified a lot about the hyperbolic stretching, thank you! My process differs from yours in that I take the image without stars to Photoshop and in addition to Camera Raw I give it a touch with Topaz Denoise IA, a comment on this would be nice. I stretch the stars in Photoshop and add them as a layer in Raster mode, it gives me better results than any other method. Siril is missing a reduction of stars like the one Bill's Script does in Pixinsight. Is there nothing comparable in Siril? Thanks again.
Nothing wrong with Topaz. I use to use it but have switched to GraXpert. And same thing with the stars. I did that previously in PS but prefer doing it in Siril now. I actually ported over one of Bill's star reduction scripts for Siril. Here's the video on that. ua-cam.com/video/Na6GzKozpCI/v-deo.html
@@DeepSpaceAstro Genius! It works very well, also on a star-only image. From what I see, it uses the first of Bill's 3 scripts, are you working on implementing the third one, which is the one I usually use? Great job! Thank you!
You know I meant to, but never got back to it. I'll have to take a look. I remember the differences in pixel math between Siril and PixInsight were a little different. That's probably why I put it on the back burner.
Your video is helpful to me. thank you I've been having trouble with GraXpert and Siril since last month, just like you. Today I got test results showing that it occurs with siril 1.2.2 and above.
I use graxpert in Windows and never found that issue. Some info, I always use 0.7 for gradient and de noise factors, and most of the times I use version 2.0 for denoise model, which is much faster than version 3 and I see not relevant improvement. In my workflow I found best results applying a second denoise in the stretched image, as the process of streching seems to reintroduce noise in the image.
I noticed that when you open Photoshop the first thing you do is adjust the black point of the three colors. Have you ever tried that in Siril's GHS. I've begun to adjust the black point of the colors separately just until they start to clip. It changes the color of the object somewhat but I kind of like it. Is this a good idea in your opinion?
Uninstalled Graxpert..Just keeps hanging when loading image. i had to keep killing the process so it would close. gave it 30 mins of messing around but i dont have time to throw away on a program that dosnt work. Thanks for your guides tho they always help as im quite new to post processing .
That's unfortunate about GraXpert. It generally works well for me although I have had it hang on occasion if I run iterative sessions with different settings on the same image. Glad to hear the videos have helped. Thanks!
Not that I'm aware of, and I'm not sure it's something that can be fixed. It may be more around the AutoStretch being too aggressive that gives it that look. When you drop to linear view and stretch, every thing looks ok.
It seems like I read somewhere that the amount of stretch applied in "auto stretch" is related to the amount of noise in the image. When you denoise the linear image the auto stretch more aggressively stretches because of the lack of noise.
That's what I thought too, but opening it back up in GraXpert has the same issue. It looks good after the denoise, it's when it's opened that things change. I don't know...
Autostretch depends on statistics of image. When you remove the gradient, the image is flatter. So the autostretch looks stronger, and the noise looks greater. But this is just an illusion.
Excellent result! I've been meaning to restack my M106 from like 4 years ago since I've learned a ton since then, You've inspired me to try it this week. Now...where did I put my raw files from 2020.
@@DeepSpaceAstro Giving it a shot now. Just discovered your channel last week and I cant begin to explain how much time you've saved from my previous workflows, but now my images are coming out how i've always wanted them
@@DeepSpaceAstro one more quick question. Do you know if the Foraxx Palette would work in Siril? I.e. is there a script that could also pull S2 out of Narrowband OSC data with the Ha and Oiii?
Did it take you several sessions to obtain the 8.4 hrs? When it came to stack them did you use Sirilc or did you just combine all the lights together in one lights folder? If using Sirilc did you have separate flats folders for each session? Did you use master dark and bias (dark flats) files? I use MacOS on a new M3 MacBook Pro with 36g of ram and found Siril 1.2.3 very slow under this platform. Apparently other MacOS users have experienced this issue. I've downgraded to 1.2.1. I've used GraxPert in 1.2.1 a couple of months ago and don't remember seeing the autostrech anomaly. Maybe try on 1.2.1.
8.4hrs? Did I say that in the video? If so, that's wrong. This was ~2.5hrs. To your question though... If I image the same object over multiple nights, and those nights were consecutive allowing me to leave my rig untouched, I use the same set of flats for all nights, and throw all the lights into the Lights folder, and stack in Siril. Nowadays I use master darks and biases, so no need to shoot those for each session. Now if those multiple nights are not consecutive, and I have to tear down and setup again each night, then yes I shoot a set of flats for each night, and then use SirilIC.
My guess on the GraXpert strange artifacts is that the data is somehow double stretched. Maybe not saving the result stretched could solve the issue? Otherwise must be something related how GX saves the resultant fit file because the result is really good.
@ I forgot to say I don’t even crop it for that purpose so to not shift . But I have to flip in the position to fit with other versions for extra layer for Ps For example: the same fit file position I view on asistudio on siril would be upside down . Than I have to flip or turn of flip option on color calibration. The stacking I do on astap because this problem I did stack calibration once then the calibration files I stacked on calibration and alinged than this two I stacked in sigma clip method and again did not help .
I tried this and found that using the human weighted luminance just stretched the image and i had to take the slider so far for anything ti come up but you could hardly see my galaxy at all, i got a better 'stretch' from doing an auto stretch in histogram transformation but like you said, its muted
@@DeepSpaceAstro so i have to do a PCC, if i dont, my image is orange instead of green after removing green noise. which i think might be why i cant use the human weighted luminance?
I have a question not associated with this video, but abour your narrowband processing and hubble pallet videos. I wanted to make a script like ExtractHaOIII but include drizzling. Currently i shoot very under sampled so drizzling is almost necessary. But I'm struggling to figure out how to do this. I read a cloudy nights thread and it didn't really help as it ended before they got it working. My other options i understand are just stacking normally, then split color channels(but that doesn't seem to give as good of an effect as the script), or possibly just stack nornally and use your automatic hubble script? What do you think? An automatic script that stacks, drizzles, and extracts into two fake mono images seems to me like the simplest route but possibly most complex to set up. Could you help me? And thank you for all your help already. I got an asi2600mc pro and upgrading to a good camera caused me to dive into seeing how to best process and you've been great help!
EDIT: I found a better way to do this vs. what I explained below. Test it out the script below and let me know So I think I have this worked out. I had to register the Ha a 2nd time so I could drizzle. The way the original script works is you end up with an image half the resolution, and then it drizzles during registration to bring it back up to full resolution. So since I had to register again to upsample, that's another sequence of files that are created, so more drive space. Probably not a big deal, but wanted to make you're aware. I've ran the modified script just a few times, and it appears to work. I'm just wondering if downsampling, then effectively drizzling twice does anything to the data. Anyway, play with it and let me know. 1drv.ms/u/s!Aii2lourR1D0gapKqjy35MW49GfXUQ?e=0Q43Eq
@@DeepSpaceAstro wow thank you! I'll definitely try it out. I was curious too if the double drizzle would affect the data or the noise at all. Thank you so much! I'll let you know how it goes.
Hey if you downloaded the script already, download it again. I somehow uploaded the 1st version that I created. The new one works more efficiently. If you open the script, the correct version is 1.1.1. sorry about that.
I don't know if this is the cause of the weirdness of Graxpret, but I was looking at the Seti Astro channel. In the "Newest Gradient Correction Tool" post, He explains how Graxpert affects the main signal. You can't see it in the first-order linear gradient. By removing that first-order gradient, you see how much Graxpert affects the signal, not just the background. I don't fully understand the process here, so I can't tell if this is the cause of the Graxpert mess. Maybe you'll be able to figure it out? This is the basis of the question I asked you yesterday about "AutoDBE". This guy does not use Graxpert. He uses AutoDBE instead.🤔
That's in Pixsight though, right? In my testing the DBE in GraXpert isn't what's causing the issue. It's only when I run the denoise that it occurs. Hoping to get to the bottom of it. I did post in their GitHub, but haven't received a response. Thanks for the info.
My sky is constantly cloudy so my learning process is taking years (even for the basics) I don't want to pay for Puxinsight yet (my first dedicated astro cam arrived today..585mc pro 😁) Siril is great and your channel really helps when I actually have data. Fantastic content. Thanks
Whenever I stack using Siril it gives me vignette and other issues which should have been removed with flats but if I use dss or other software it does it correctly. Any idea?
It may be your data. If you're using the script, it can include subs that may not be that good, and maybe DSS is a little more strict with it's settings. Either go through the subs and remove any that are bad, i.e. clouds, you don't use the script and stack manually adjusting the settings as needed.
I think it must use the SNR somehow to know how far to strech the autostretch. It stretches it more after GraXpert. I like the difference actually. It allows for better manual color calibration almost all the time for me. I never use PCC. Funny.... your workflow is nearly identical to mine.
I don't think so. The issue is when viewing in Autostretch, and those sliders aren't available in that view. I did hear from the dev and they said it's because the MTF is only for the current session of GraXpert. It's not saved like it is for the .xisf for PixInsight.
That's where I leave it. It actually appears to be just for the view in GraXpert. It doesn't save the saturation adjustment, and if you hover you mouse over the Save Stretched & Processed button it says that the color saturation is not changed.
@@DeepSpaceAstro I mostly just choose the save processed because during the siril if the second is chosen it messes up the stretching im still confused what the hover for processed means
Like always great video . Question on siril why don’t you search for the object whe you do color calibration? When I do background extraction on Graxpert looks like bring more noise to it but the same image I take in to Siril then it looks fine I experiment to denoise on siril after Star removal siril did denoise as good as graxpert to my opinion .
You only need to search for the object if the information needed is not in the FITS header of the images, like if you're imaging with an DSLR. I'm using an astronomy camera and NINA writes that data to the header.
@@DeepSpaceAstro I did try both ways I did search and not search the result came the same I’m using zwoasi 2600 mc pro looks like you said it maight have the information.
cool video, just one comment. If you use photometric color calibration, you need to give it an object from the catalog in the "find" field, otherwise the whole point of photometrics is void :)
I desaturate the stars before deconvolution though. That works pretty well. I also use the highlight protection too a lot to protect the brightest parts.
It's in your working directory with the starless image. If it's not, then maybe you didn't have the create starmask option checked? Nothing is ruined. Your original stack is still in your working directory as well.
@@DeepSpaceAstro You talk about processing using the scripts, why do you prefer the scripts over SiriLIC? Trying to figure out if I’m missing features?
Siril already has a background extraction tool, why do you complicate your life with yet another software?, genuine question here, I want to understand whats the reasoning behind this extra step
@@DeepSpaceAstro Still is really annoying because I have no idea what im looking at it messes up with the nebulosity and im confused which was the right orientation Also graxpert loves to crash a lot.
Awesome vide I had to fallow entire vide and I did excellent job from you know The only one I had to skip is deconvolution because on me wasn’t working right Another version I ad some saturation stretch on GHST AND SOME On color saturation. I wish I can share my pictures some where . Thank you so much .
I run astap and graxpert but sirli did not give me the same trouble as you had with deconvolution but I couldn’t remove the circle around the stars after clicking apply How do I remvome the selected stars circle ? Thanks
@@DeepSpaceAstro I’m sorry for not being clear when I go to dynamic PSF I SELECT THOSE 3 yellow stars below And the image has selected some stars with pink circle or pink lines others look orange to me I close dynamic PSF than I go to deconvolution I select generate PSF after I choose needed parameters . Now when I apply the result . What I have selected on Dinamic PSF is not going a way is still there . On your video they go a way
@@DeepSpaceAstro What I found out to keep default setting on iterations and gredient descent step size it turned all green pixels around m31 now if I try your setting on that now it tuned all over pink pixels or dots and orange circles reminded on the selected stars .
It’s so confusing with astap Graxpert and Siri and Asi studio ni relation In Asi studio file from astap looks beautiful From Graxpert blue sky in background after de noise and background check. The same file on siril astap green , Graxpert just fine Now when I do background extraction and green removal in siril with atap file is just fine
I'm new to siril and this has become my goto channel for any issues and techniques I need to know. Awesome work my friend .
Thank you!
Really this is a fantastic resource. I started working in Siril because I can’t justify the cost of pixinsight.
Can't tell you how much I appreciate your tutorials. I'm also fairly new to it and every time I get a little lost I go back to these as references. I don't seem to have the issue you're having between Graxpert and Siril though. Not sure why but mine looked the same in Siril as I left them in Graxpert. Thanks again. I don't subscribe to many channels, but you've earned this guy's click.
Hey thanks for the comment! That's interesting that you don't see the same. Do you use the denoise, or just the BGE by chance?
@@DeepSpaceAstro I used both. I also had just installed ver 1.2.3 after I had some issues plate solving. Don't know if it played a role or not.
Fantastic! Delighted with the results after following this. Excellent work
Great to hear! Thanks!
A great ‘refresher’ video for me Rich. I’ve been using Siril and Graxpert for a while now after watching your earlier tutorials and I endorse what you say early on about how the unstretched image looks very odd! I don’t know the answer.
I’ve never got to terms with ‘Deconvolution’ and must now give that a try. Thanks also for the tips you include about GHS and how best to preserve colour. I’ve got to give that a try too.
Thanks as always 🤝
Thank you!
I do the denoise in Graxpert AFTER I stretch the image. It's a pain jumping back and forth but o well.
Yeah that's probably an easier way to go about it. Too bad right now we can't denoise while still linear. Pretty sure they recommend denoising before stretching, but like you said, it's a pain.
I'll need to do that also. Thanks.
@DeepSpaceAstro I didn't know to denoise before stretching. Learned something new.
I too didn't have this problem with the latest version of Graxpert.
However, I could not plate solve the saved image in Siril, so I colour calibrated and removed green noise before using Graxpert. Not sure if that makes a difference.
I only used 10 x 300s exposures on the lobster nebula, so there was a lot of noise. Graxpert did an excellent job removing this. Thanks for this.
UPDATE:
I got better results applying Graxpert to the starless image (without cropping), processing in Photoshop, then recombining in Siril.
Thanks!
Thank you!
This is awesome! I can’t wait to use this process on some of my data. Thanks a lot for sharing. :)
Hope it's helpful for you! Thanks!
@@DeepSpaceAstro I have used this workflow on three sets of data so far and it has been AMAZINGLY helpful. I have been able to stretch out so much detail I didn't even realize was there when I tried processing this stuff a handful of months ago. Thanks so much.
@@jessfritsche Awesome! Glad to hear that!
Funnily enough your workflow matches mine. That's because I didn't have one and now I do!
Your channel is a Godsend, much appreciated and keep 'em coming!
😂 Thanks! Happy to hear the videos have helped out!
Questions/comments:
I've noticed the deconvolution problems but hadn't attributed them to GraXpert. With small Seestar images, I find cropping is not necessary. Process is fast. Thanks for tip on reducing no. of iterations. In adjusting step size, try initial step increase to .0004 or .0005 to save time.
In another video, you said AstroSharp may conflict with deconvolution. Any update?
Nice tip about color stretch model. Is that true for nebulae too? I notice you did not use symmetry pt. for first 2 stretches.
In linear stretch, you can click on histogram to set black point. Towards the end of processing, I try to make black area ~7-10% saturation.
I noticed you did not use Highlight protection for galaxy core & didn't check the saturation there. I've been trying to avoid overstretching of bright areas (and bright stars) when I stretch them.
In final image, sometimes stars have a green cast (perhaps due to independant color stretch?), so I use remove green noise tool. Also, I often increase saturation and/or do a saturation stretch rather than using my old (CS4) version of Photoshop.
Sometimes after Siril processing, I use another noise reduction in GraXpert and/or the last two noise reduction processes in Siril. But occasionally these introduce serious artifacts in a galaxy core. Perhaps from GraXpert?
It's not just AstroSharp, you probably shouldn't combine any sharpening tools together, but you can always try and see what happens.
Regarding the rest of your comments, there's not one way to do anything. Experiment with the tools to get the results you're happy with.
Just wow, great video, explaining everything as you go along, who could ask for more.
Glad you liked it! Thanks!
That’s helped me no end especially with the stretch function, I was deffo struggling with that, thanks for another great vid 👍
Glad it helped! Thanks!
Same problem with graxpert. Dont remember it occuring in previous version
You figure something out with graxc?
@@hmuphilly9129 graxpert is mostly ok for me at the moment. I get occasional issues but like more a problem with my data. Its invaluable
Hello, I'm using Linux Version of both software. No strange stretch problem when I re-open fits in Siril.
It seems to works great!
After using GrapXpert I have 6420 stars in my image instead of 2300 (using only Siril)
Hmmm. that's interesting.
Well, I know what I'll be watching over my lunch break. Siril is a powerful package, but it's so underused in my worflow since I only use it for stacking, background normalisation and removing green noise. I should really do a deep dive into the documentation one of these days...
If you haven't seen it, I have a start to finish video to get you started. ua-cam.com/video/9K-V2VIcwfQ/v-deo.html
@@DeepSpaceAstro Yup, I noticed you mentioned it in this video, so I'm already watching it. So far, it looks more or less like the process I'm already following, which I got from Nico over at Nebula Photo's. Except from what I gather, you're not leaving Siril to post-process in Gimp or Photoshop. So curious to see what that workflow looks like.
I do use PS to post process. I show that towards the end of that video.
Im curious what you might do different for pure clusters, like opens. I’m shooting at 1950efl and I’ve gotten some good globular details but I haven’t really landed on a foolproof fundamental approach. Starnet+ is a core part of my Galaxy/nebula workflow, and in Recomposition my goal is to keep the star fields from overwhelming the subject. But for clusters, the stars are the show. Much tougher I find.
Try this workflow. ua-cam.com/video/xMSQWHqLzmU/v-deo.html
@@DeepSpaceAstro I love it. Very similar to my basic workflow but I note the differences, and they make good sense. I wonder if those 'galaxies' could be saved by doing a stretch, then taking that starless into PS as a tif and with Color Range Select Shadows getting a mask over the blown-out core and sliding the Exposure down, re-saving into Siril as a fit then using that for a Recomposition. I suppose that could create a nasty artifact the other way though?
Thank you Rich, I've learned so much with your videos !
You're welcome! Glad to hear they've been helpful!
Hey Rich, thanks for sharing. Not having read all the comments - have you tried using Graxpert from within Siril? I don’t get that “weird stretching” when I use it that way. Also I’m using the alpha-dev 1.3.2 version of Siril.
I've seen it, but have played with it yet. Good to hear we won't have this issue in the next version.
I ve learn Siril because of you. Thanks again I owe you😊
Happy to hear that! Thanks!
This is great. You are great. Subscribed!
Awesome, thank you!
Wow excellent video
Excellent result too .
Thanks
Thanks!
Hey Rich , thanks so much for this. I took me a while but I made it through. Phew
Glad it helped!
Hey, Rich, thanks for another on-point video tutorial. I have also been experiencing some WEIRD side effects when hopping back and forth between Siril and GraXpert. There seems to be some kind of mismatch between the two programs. Most recently, I noted it when processing a wide-angle image of the Veil Nebula complex with that dense star field surrounding it. I am not sure if the problem is directly related to the order of operations or what. I was trying to use the AI versions of GraXpert gradient removal and denoising. Weird.
Yeah it's the changes it makes to the linear data that makes the Autostretch too much afterwards. The data itself is ok. For now we just have to deal with it.
...After Graxpert I get back to Siril and stay only in the "Line- Mode", stretching my image there very lightly with the Asinh Transformation. An other good way is the Arcsinh- Curves- adjustment- workflow at Photoshop. After all I go ahead as I learn from you: PSF...Deconvolution... Starnet++ and so on. Greetings from Thailand
I had never tried the Asinh Transformation. It makes stretching so simple. Thank you!
Great video my friend,I always use siril for astro and your tutorials are always very helpful. I've got a 130mm reflector I use for astro but have been unable to get a decent image of a globular, a tutorial on capture and processing in siril would be super useful if possible... many thanks
Thanks! Have you seen my beginners processing tutorial?
@@DeepSpaceAstro I have my friend. Can't remember a globular though?
It would be pretty much the same workforce, but skip removing the stars.
Thanks. I'll check it out again 😃
Hi Rich, thank you for your channel as it is really very helpful to me. But I have a little problem and I would like to make some points. The moment I recompose the starless and starmask files, and save in Tiff the final file, I notice some loss of detail and color. So I thought of this solution, which at least works for me. Considering that I don't trust the star recomposition process nor Siril's export process, I open the two Fit files with GraXpert, export them to Tiff and work them in Photoshop creating two layers and applying screens to the star layer and...boom...the image is now as perfect as I expect !
Definitely a valid way to do it. That's pretty much the process before we had Starnet integration in Siril.
Hello Rich, your video is very good!
It's more or less how I've been working, you clarified a lot about the hyperbolic stretching, thank you!
My process differs from yours in that I take the image without stars to Photoshop and in addition to Camera Raw I give it a touch with Topaz Denoise IA, a comment on this would be nice.
I stretch the stars in Photoshop and add them as a layer in Raster mode, it gives me better results than any other method.
Siril is missing a reduction of stars like the one Bill's Script does in Pixinsight. Is there nothing comparable in Siril?
Thanks again.
Nothing wrong with Topaz. I use to use it but have switched to GraXpert. And same thing with the stars. I did that previously in PS but prefer doing it in Siril now. I actually ported over one of Bill's star reduction scripts for Siril. Here's the video on that. ua-cam.com/video/Na6GzKozpCI/v-deo.html
@@DeepSpaceAstro Genius! It works very well, also on a star-only image.
From what I see, it uses the first of Bill's 3 scripts, are you working on implementing the third one, which is the one I usually use?
Great job! Thank you!
You know I meant to, but never got back to it. I'll have to take a look. I remember the differences in pixel math between Siril and PixInsight were a little different. That's probably why I put it on the back burner.
Your video is helpful to me. thank you
I've been having trouble with GraXpert and Siril since last month, just like you.
Today I got test results showing that it occurs with siril 1.2.2 and above.
Maybe I thought wrong.
I'll test it a little more.
I use graxpert in Windows and never found that issue. Some info, I always use 0.7 for gradient and de noise factors, and most of the times I use version 2.0 for denoise model, which is much faster than version 3 and I see not relevant improvement.
In my workflow I found best results applying a second denoise in the stretched image, as the process of streching seems to reintroduce noise in the image.
I noticed that when you open Photoshop the first thing you do is adjust the black point of the three colors. Have you ever tried that in Siril's GHS. I've begun to adjust the black point of the colors separately just until they start to clip. It changes the color of the object somewhat but I kind of like it. Is this a good idea in your opinion?
It's all preference at this point. If you like the way it looks, then it's right. 😉
Uninstalled Graxpert..Just keeps hanging when loading image. i had to keep killing the process so it would close. gave it 30 mins of messing around but i dont have time to throw away on a program that dosnt work. Thanks for your guides tho they always help as im quite new to post processing .
That's unfortunate about GraXpert. It generally works well for me although I have had it hang on occasion if I run iterative sessions with different settings on the same image. Glad to hear the videos have helped. Thanks!
Hi Rich, has the Graxpert issue with autostrech been resolved?
Not that I'm aware of, and I'm not sure it's something that can be fixed. It may be more around the AutoStretch being too aggressive that gives it that look. When you drop to linear view and stretch, every thing looks ok.
Thanks for getting back to me! Love your observatory! What is your "day job"?
Thanks! I'm a tech consultant.
It seems like I read somewhere that the amount of stretch applied in "auto stretch" is related to the amount of noise in the image. When you denoise the linear image the auto stretch more aggressively stretches because of the lack of noise.
That's what I thought too, but opening it back up in GraXpert has the same issue. It looks good after the denoise, it's when it's opened that things change. I don't know...
Autostretch depends on statistics of image. When you remove the gradient, the image is flatter. So the autostretch looks stronger, and the noise looks greater. But this is just an illusion.
But I don't see that happening with background extraction. Only after I run denoise.
@@DeepSpaceAstro probably because denoise changes a bit the dynamic.
So use the HD version of auto stretch
Excellent result! I've been meaning to restack my M106 from like 4 years ago since I've learned a ton since then, You've inspired me to try it this week.
Now...where did I put my raw files from 2020.
Thanks Naz! Look in your Covid folder. 😜
If i'm using the Extract Ha and Oiii script for stacking, should i run both of those result images in GraXpert for Denoise before combining?
You can do it either way. Try both and see what works best.
@@DeepSpaceAstro Giving it a shot now. Just discovered your channel last week and I cant begin to explain how much time you've saved from my previous workflows, but now my images are coming out how i've always wanted them
@@DeepSpaceAstro one more quick question. Do you know if the Foraxx Palette would work in Siril? I.e. is there a script that could also pull S2 out of Narrowband OSC data with the Ha and Oiii?
@@GuitarKid581 I don't think so. SII is too close to Ha to pull it from NB
What was the exposure and final integration time for your M106 data? Thanks.
101x300" exposures (8.4 hrs)
Did it take you several sessions to obtain the 8.4 hrs? When it came to stack them did you use Sirilc or did you just combine all the lights together in one lights folder? If using Sirilc did you have separate flats folders for each session? Did you use master dark and bias (dark flats) files? I use MacOS on a new M3 MacBook Pro with 36g of ram and found Siril 1.2.3 very slow under this platform. Apparently other MacOS users have experienced this issue. I've downgraded to 1.2.1. I've used GraxPert in 1.2.1 a couple of months ago and don't remember seeing the autostrech anomaly. Maybe try on 1.2.1.
8.4hrs? Did I say that in the video? If so, that's wrong. This was ~2.5hrs. To your question though... If I image the same object over multiple nights, and those nights were consecutive allowing me to leave my rig untouched, I use the same set of flats for all nights, and throw all the lights into the Lights folder, and stack in Siril. Nowadays I use master darks and biases, so no need to shoot those for each session.
Now if those multiple nights are not consecutive, and I have to tear down and setup again each night, then yes I shoot a set of flats for each night, and then use SirilIC.
My guess on the GraXpert strange artifacts is that the data is somehow double stretched. Maybe not saving the result stretched could solve the issue? Otherwise must be something related how GX saves the resultant fit file because the result is really good.
I don't save the stretched result. It's more around what it does to the linear data that makes Siril's AutoStretch too aggressive.
Hi again
My starless don’t fit right in line position with starmask as they where before starnet ++
What can be wrong ?
If you cropped either image after running Starnet, that would cause it not to be able to recompose the image.
@
I forgot to say I don’t even crop it for that purpose so to not shift .
But I have to flip in the position to fit with other versions for extra layer for Ps
For example: the same fit file position I view on asistudio on siril would be upside down . Than I have to flip or turn of flip option on color calibration.
The stacking I do on astap because this problem I did stack calibration once then the calibration files I stacked on calibration and alinged than this two I stacked in sigma clip method and again did not help .
I tried this and found that using the human weighted luminance just stretched the image and i had to take the slider so far for anything ti come up but you could hardly see my galaxy at all, i got a better 'stretch' from doing an auto stretch in histogram transformation but like you said, its muted
Try GHS again after using the histogram transformation.
@@DeepSpaceAstro so i have to do a PCC, if i dont, my image is orange instead of green after removing green noise. which i think might be why i cant use the human weighted luminance?
I have a question not associated with this video, but abour your narrowband processing and hubble pallet videos. I wanted to make a script like ExtractHaOIII but include drizzling. Currently i shoot very under sampled so drizzling is almost necessary. But I'm struggling to figure out how to do this. I read a cloudy nights thread and it didn't really help as it ended before they got it working. My other options i understand are just stacking normally, then split color channels(but that doesn't seem to give as good of an effect as the script), or possibly just stack nornally and use your automatic hubble script? What do you think? An automatic script that stacks, drizzles, and extracts into two fake mono images seems to me like the simplest route but possibly most complex to set up. Could you help me? And thank you for all your help already. I got an asi2600mc pro and upgrading to a good camera caused me to dive into seeing how to best process and you've been great help!
EDIT: I found a better way to do this vs. what I explained below. Test it out the script below and let me know
So I think I have this worked out. I had to register the Ha a 2nd time so I could drizzle. The way the original script works is you end up with an image half the resolution, and then it drizzles during registration to bring it back up to full resolution. So since I had to register again to upsample, that's another sequence of files that are created, so more drive space. Probably not a big deal, but wanted to make you're aware. I've ran the modified script just a few times, and it appears to work. I'm just wondering if downsampling, then effectively drizzling twice does anything to the data. Anyway, play with it and let me know. 1drv.ms/u/s!Aii2lourR1D0gapKqjy35MW49GfXUQ?e=0Q43Eq
@@DeepSpaceAstro wow thank you! I'll definitely try it out. I was curious too if the double drizzle would affect the data or the noise at all. Thank you so much! I'll let you know how it goes.
Hey if you downloaded the script already, download it again. I somehow uploaded the 1st version that I created. The new one works more efficiently. If you open the script, the correct version is 1.1.1. sorry about that.
@@DeepSpaceAstro no worries! Thanks again for your expertise and willingness to help!
I don't know if this is the cause of the weirdness of Graxpret, but I was looking at the Seti Astro channel. In the "Newest Gradient Correction Tool" post, He explains how Graxpert affects the main signal. You can't see it in the first-order linear gradient. By removing that first-order gradient, you see how much Graxpert affects the signal, not just the background. I don't fully understand the process here, so I can't tell if this is the cause of the Graxpert mess. Maybe you'll be able to figure it out? This is the basis of the question I asked you yesterday about "AutoDBE". This guy does not use Graxpert. He uses AutoDBE instead.🤔
That's in Pixsight though, right? In my testing the DBE in GraXpert isn't what's causing the issue. It's only when I run the denoise that it occurs. Hoping to get to the bottom of it. I did post in their GitHub, but haven't received a response. Thanks for the info.
@@DeepSpaceAstro Thanks for the clarification.
Thx for sharing...greets from germany
@@DeepSpaceAstro wonder if you would get any response from the graxpert discord.
Thank you so much for this video
Glad it was helpful! Thanks!
My sky is constantly cloudy so my learning process is taking years (even for the basics) I don't want to pay for Puxinsight yet (my first dedicated astro cam arrived today..585mc pro 😁) Siril is great and your channel really helps when I actually have data. Fantastic content. Thanks
Thanks so much!
Thank you , you pointed out what I've been doing wrong in my processing.
Happy to hear it was helpful!
It looks better when I use the Asinh stretch instead of the auto stretch mode.
Whenever I stack using Siril it gives me vignette and other issues which should have been removed with flats but if I use dss or other software it does it correctly. Any idea?
It may be your data. If you're using the script, it can include subs that may not be that good, and maybe DSS is a little more strict with it's settings. Either go through the subs and remove any that are bad, i.e. clouds, you don't use the script and stack manually adjusting the settings as needed.
@@DeepSpaceAstro I’ll try
I think it must use the SNR somehow to know how far to strech the autostretch. It stretches it more after GraXpert. I like the difference actually. It allows for better manual color calibration almost all the time for me. I never use PCC.
Funny.... your workflow is nearly identical to mine.
game changer!!!
Thanks!
Thanks for the guide
Happy to help!
Are you now preferring to use the Camera Raw clarity slider for sharpening rather than the Siril wavelets tool?
I still use it on occasion.
maybe because of one of the slider below in siri window? I always get max value, not 59195.
I don't think so. The issue is when viewing in Autostretch, and those sliders aren't available in that view. I did hear from the dev and they said it's because the MTF is only for the current session of GraXpert. It's not saved like it is for the .xisf for PixInsight.
Do you want the graxpert saturation at 1.0?
That's where I leave it. It actually appears to be just for the view in GraXpert. It doesn't save the saturation adjustment, and if you hover you mouse over the Save Stretched & Processed button it says that the color saturation is not changed.
@@DeepSpaceAstro I mostly just choose the save processed because during the siril if the second is chosen it messes up the stretching
im still confused what the hover for processed means
Like always great video .
Question on siril why don’t you search for the object whe you do color calibration?
When I do background extraction on Graxpert looks like bring more noise to it but the same image I take in to Siril then it looks fine
I experiment to denoise on siril after Star removal siril did denoise as good as graxpert to my opinion .
You only need to search for the object if the information needed is not in the FITS header of the images, like if you're imaging with an DSLR. I'm using an astronomy camera and NINA writes that data to the header.
@@DeepSpaceAstro
I did try both ways I did search and not search the result came the same
I’m using zwoasi 2600 mc pro looks like you said it maight have the information.
cool video, just one comment. If you use photometric color calibration, you need to give it an object from the catalog in the "find" field, otherwise the whole point of photometrics is void :)
You only need to do that if the RA/DEC data of the object doesn't exist in the FITS header.
I desaturate the stars before deconvolution though. That works pretty well.
I also use the highlight protection too a lot to protect the brightest parts.
Good tip ! Thanks!
couldnt find my starmask file and it ruined the whole 2 hour process :(
It's in your working directory with the starless image. If it's not, then maybe you didn't have the create starmask option checked? Nothing is ruined. Your original stack is still in your working directory as well.
i have that issue too
I never see any mention of SiriLIC in anyone’s videos. Is anyone other than me using it?
Rick made a video on that about 1 year ago
ua-cam.com/video/HJHz1dEy62c/v-deo.htmlsi=4rh63HFbZidZCG42
We sure are, and I have a video on it as well. ua-cam.com/video/HJHz1dEy62c/v-deo.html
@@DeepSpaceAstro You talk about processing using the scripts, why do you prefer the scripts over SiriLIC? Trying to figure out if I’m missing features?
The scripts generally work well. I only manually stack and process if the data requires it, i.e. can't register some images, etc.
ive started using it recently, one thing i havent figured out without editing the project json is adding on additional sessions
Siril already has a background extraction tool, why do you complicate your life with yet another software?, genuine question here, I want to understand whats the reasoning behind this extra step
I like the AI model in GraXpert.
очень полезное видео!
Thanks!
Wonder if the graxpert devs are aware..
I posted in their github, and the one of the Siril devs reached out to them on my behalf as well about it. Not sure what will come of it.
@@DeepSpaceAstro Still is really annoying because I have no idea what im looking at it messes up with the nebulosity and im confused which was the right orientation
Also graxpert loves to crash a lot.
Sorry for being so curious, I'm a BMW engineer😊
Awesome vide I had to fallow entire vide and I did excellent job from you know
The only one I had to skip is deconvolution because on me wasn’t working right
Another version I ad some saturation stretch on GHST AND SOME On color saturation.
I wish I can share my pictures some where .
Thank you so much .
Instagram and Facebook links in the description. You can tag me there
I run astap and graxpert but sirli did not give me the same trouble as you had with deconvolution but I couldn’t remove the circle around the stars after clicking apply
How do I remvome the selected stars circle ?
Thanks
Sorry? Stars circle?
@@DeepSpaceAstro
I’m sorry for not being clear when I go to dynamic PSF I SELECT THOSE 3 yellow stars below
And the image has selected some stars with pink circle or pink lines others look orange to me I close dynamic PSF than I go to deconvolution I select generate PSF after I choose needed parameters .
Now when I apply the result .
What I have selected on Dinamic PSF is not going a way is still there .
On your video they go a way
Maybe this will help? ua-cam.com/video/dC5hyDeGhKk/v-deo.html
@@DeepSpaceAstro
What I found out to keep default setting on iterations and gredient descent step size it turned all green pixels around m31 now if I try your setting on that now it tuned all over pink pixels or dots and orange circles reminded on the selected stars .
@@DeepSpaceAstro let me try this
Thanks .
It’s so confusing with astap Graxpert and Siri and Asi studio ni relation
In Asi studio file from astap looks beautiful
From Graxpert blue sky in background after de noise and background check.
The same file on siril astap green , Graxpert just fine
Now when I do background extraction and green removal in siril with atap file is just fine