I think fundamentally it boils down to good idea, bad implementation. They are trying to match a mmt of a short focal length reference image to a mmt of the long focal length target image. Subtract them, then ideally you are just left with gradient. However since the target image is necessarily going to have more detail and SNR the mmt's will never match. The subtraction will strip a ton of signal. What they should do is use something like ADBE on the target image. Use ADBE on the reference image (assuming it is already "flat"). Subtract those two backgrounds. Now that is the real gradient that needs to be removed from the original target image. They dont make the reference data accessible otherwise we could implement something like that
Yep, they took the scientific rigor right out of it as soon as they said you have to eyeball your results. Maybe I shouldve been harder in my review, but it is version 1 release and I know my version 1's get changed almost immediately. I even had a chance to play with it for a week before the public release and can see all the fiddling I did in my video with it still.
@@chrisbeere2117 i think fundamentally it boils down to good idea, bad implementation. They are trying to match a mmt of a short focal length reference image to a mmt of the long focal length target image. Subtract them, then ideally you are just left with gradient. However since the target image is necessarily going to have more detail and SNR the mmt's will never match. To the subtraction will strip a ton of signal. What they should do is use something like ADBE on the target image. Use ADBE on the reference image. Subtract those two backgrounds. Now that is the real gradient that needs to be removed from the original target image. They dont make the reference data accessible otherwise we could implement something like that
Thanks for the in depth review of the pros and cons and the quick walk thru of the options (sliders) for adjustments. It's more complicated than ADBE which I will continue to use ADBE
My heart sank when I saw that you need to run Image Solver, it has never worked very well for me. Then I remembered you have a Blind Solver script and now I’m excited to give it a test drive, so thanks for that!
You mentioned several times that certain areas are over corrected, or under corrected. From your perspective, i gather that it's a fairly simple determination. However, I have no reference for what a properly-corrected area of nebulosity SHOULD look like. Do you have any guidance on how to develop the skill to quickly determining whether an area is over- or under-corrected? Perhaps a video?
See that is part of tge MGC tool. It loses all scientific rigor once people have to eyeball this stuff. You can see others that are saying it is nuking their nebula as well. Anytime you see something like that they are saying overcorrecting. Removing signal and not gradoent
Thanks for this video. I think I have a pretty good handle now how to use it. I too thought when Mars came out, it would be pretty much a button push. Even though it’s not, it still is awesome as a tool.
I'm still mostly pulling gradients manually. True, I only have to deal with gradients created by moonlight, but I find I get better results just extracting them out through application of discreet tools in Affinity Photo and PhotoLab 8.
You arent missing anything here. Listening to the community out there my ADBE and grax are still the go to. PI lost all their scientific rigor with this as soon as they staryed having the user eyeball the gradient for best results. Why even use the flux calibration then. Just jump straight to messing with scale amounts. I am pretty disappointed to say the least.
Mgc doesnt create mosaics. It just removes gradients. You would do this first prior to making your mosaic. I already have a mosaic tutorial which still is accurate to v1.9 currently
nice, between this and pixinsight’s MGC optimization video i feel like i’ll have a better idea of how to adjust things to get the best results. i too am a bit surprised that there’s less scientific rigour and more ‘expectations’ involved in getting a good result, but it makes sense. I’d also like to see how this goes in using your own widefield image taken at the same time, like if you’re imaging something that’s outside of the reference survey right now. as well as, like what many people are already saying, how well this will work for OSC dual band filter setups
yes, they start out strong with flux calibrations and gaia data and all that, then kind of through the rigor out the window and need the user to modify it all based on how they feel it should be looking. hoping it gets better as it improves.
Thanks for the video Frank, unfortunately us Southerners have to wait for the database to be updated as it only features some southern , most Milkyway and the Northern skies. Would love you to work out a script to simplify the process. That would be awesome.
It does seem extra convoluted. Like they couldve combined this a little. Tons of work just to get to use MGC, then there is a good chance you are just playing with sliders again and eyeballing the results. Overall a bit disappointing tbh considering they are saying this is now THE gradient removal solution
Yeah a bit for sure. Even in the manual process laid out by Vicet there is a k value you have to adjust manually to get the good removal. Reminds me of continuum subtraction before nightphotons figured out better methods that we were then able to automate exactly
Good first video. But what do the fitler selection dropdowns in MGC do? How do they relate to the RGB channels in my image? I am using an OSC with an Askar D1 filter, so my channels are not really pure Ha and Oiii data. I have been leaving them as defaults (R, G and B). In the flux calculation I have been choosing the Askar filters, but for QE curve I have been using "Ideal QE curve" as this is what is suggested in SPCC documentation.
For spfc you will use the narrowband check. For mgc use the narrowband dropdowns. It is as close as you can get. If you know your sensor you should be using that instead of ideal qe as well
Actually i am not sure it matters a whole lot. You are already going to be moving scale factors, structure seperation, etc anyways and eyeballing it, the scientific rigor is right out the window anyways
I think fundamentally it boils down to good idea, bad implementation. They are trying to match a mmt of a short focal length reference image to a mmt of the long focal length target image. Subtract them, then ideally you are just left with gradient. However since the target image is necessarily going to have more detail and SNR the mmt's will never match. The subtraction will strip a ton of signal. What they should do is use something like ADBE on the target image. Use ADBE on the reference image (assuming it is already "flat"). Subtract those two backgrounds. Now that is the real gradient that needs to be removed from the original target image. They dont make the reference data accessible otherwise we could implement something like that
Honestly after testing it and seeing how it still nuked some nebulosity, and required a lot of operator skill... I'm still using your ADBE and GXP...
Yep, they took the scientific rigor right out of it as soon as they said you have to eyeball your results. Maybe I shouldve been harder in my review, but it is version 1 release and I know my version 1's get changed almost immediately. I even had a chance to play with it for a week before the public release and can see all the fiddling I did in my video with it still.
Yeah i have the exact same problem, it rips out a ton of signal.
@@chrisbeere2117 i think fundamentally it boils down to good idea, bad implementation. They are trying to match a mmt of a short focal length reference image to a mmt of the long focal length target image. Subtract them, then ideally you are just left with gradient. However since the target image is necessarily going to have more detail and SNR the mmt's will never match. To the subtraction will strip a ton of signal. What they should do is use something like ADBE on the target image. Use ADBE on the reference image. Subtract those two backgrounds. Now that is the real gradient that needs to be removed from the original target image. They dont make the reference data accessible otherwise we could implement something like that
you are such a great and kind chap Fanklin! THankyou
Thanks for the in depth review of the pros and cons and the quick walk thru of the options (sliders) for adjustments. It's more complicated than ADBE which I will continue to use ADBE
Sounds like a lot of people will continue to do so until more rigor gets into this tool
Most important take-away that will be overlooked by many....it's NOT a magic wand. Excellent presentation.
My heart sank when I saw that you need to run Image Solver, it has never worked very well for me. Then I remembered you have a Blind Solver script and now I’m excited to give it a test drive, so thanks for that!
You mentioned several times that certain areas are over corrected, or under corrected. From your perspective, i gather that it's a fairly simple determination.
However, I have no reference for what a properly-corrected area of nebulosity SHOULD look like.
Do you have any guidance on how to develop the skill to quickly determining whether an area is over- or under-corrected?
Perhaps a video?
See that is part of tge MGC tool. It loses all scientific rigor once people have to eyeball this stuff. You can see others that are saying it is nuking their nebula as well. Anytime you see something like that they are saying overcorrecting. Removing signal and not gradoent
Thank you Frank …and Merry Christmas and Happy Holidays to you and family.😊❤
Thank you very much for the video. It gives us some guidance as we try to progress in astrophotography. Greetings from Turkey.
Thanks a lot for this. Really helps clarify a lot of stuff around this new future.
I think that was a fair and balanced analysis of the first release. Thank you!
Thanks for this video. I think I have a pretty good handle now how to use it. I too thought when Mars came out, it would be pretty much a button push. Even though it’s not, it still is awesome as a tool.
Yep for now need to consider it another gradient removal tool to try as it is not fully there yet imho
I'm still mostly pulling gradients manually. True, I only have to deal with gradients created by moonlight, but I find I get better results just extracting them out through application of discreet tools in Affinity Photo and PhotoLab 8.
You arent missing anything here. Listening to the community out there my ADBE and grax are still the go to. PI lost all their scientific rigor with this as soon as they staryed having the user eyeball the gradient for best results. Why even use the flux calibration then. Just jump straight to messing with scale amounts. I am pretty disappointed to say the least.
Many thanks for this video. I think I'll stick with DBE for the time being.
Could you demonstrate how to use MGC to create a mosaic?
Mgc doesnt create mosaics. It just removes gradients. You would do this first prior to making your mosaic. I already have a mosaic tutorial which still is accurate to v1.9 currently
nice, between this and pixinsight’s MGC optimization video i feel like i’ll have a better idea of how to adjust things to get the best results. i too am a bit surprised that there’s less scientific rigour and more ‘expectations’ involved in getting a good result, but it makes sense. I’d also like to see how this goes in using your own widefield image taken at the same time, like if you’re imaging something that’s outside of the reference survey right now. as well as, like what many people are already saying, how well this will work for OSC dual band filter setups
yes, they start out strong with flux calibrations and gaia data and all that, then kind of through the rigor out the window and need the user to modify it all based on how they feel it should be looking. hoping it gets better as it improves.
ADBE is my goto. Absolutely love the one push button
Doesnt seem like MGC is as rigorous as we were hoping for sure and just another fiddly gradient removal tool at the moment
Thanks for the video Frank, unfortunately us Southerners have to wait for the database to be updated as it only features some southern , most Milkyway and the Northern skies. Would love you to work out a script to simplify the process. That would be awesome.
It does seem extra convoluted. Like they couldve combined this a little. Tons of work just to get to use MGC, then there is a good chance you are just playing with sliders again and eyeballing the results. Overall a bit disappointing tbh considering they are saying this is now THE gradient removal solution
@@setiv2 I just feel they released it too early and it's just their gradient correction tool dressed up in new clothing
Yeah a bit for sure. Even in the manual process laid out by Vicet there is a k value you have to adjust manually to get the good removal. Reminds me of continuum subtraction before nightphotons figured out better methods that we were then able to automate exactly
Awesome, thanks
Good first video. But what do the fitler selection dropdowns in MGC do? How do they relate to the RGB channels in my image? I am using an OSC with an Askar D1 filter, so my channels are not really pure Ha and Oiii data. I have been leaving them as defaults (R, G and B). In the flux calculation I have been choosing the Askar filters, but for QE curve I have been using "Ideal QE curve" as this is what is suggested in SPCC documentation.
For spfc you will use the narrowband check. For mgc use the narrowband dropdowns. It is as close as you can get. If you know your sensor you should be using that instead of ideal qe as well
Actually i am not sure it matters a whole lot. You are already going to be moving scale factors, structure seperation, etc anyways and eyeballing it, the scientific rigor is right out the window anyways
Not sure why anyone would use this. DBE is easy, much less work than this, for the results you can get.
Use my ADBE, even easier :)