First attempts at using Pixinsight

[mightymonoped]

Finally succumbed to integrating PI into my toolset. 
Already use the sublime StarXterminator and NoiseXterminator too, but finally get to add BlurXterminator to the fold. 
 

Started things off with a few reprocesses of old HOO captures, converting to SHO, following Lukomatico’s excellent tutorials. 
 

First steps in a long journey I think but looking forward to using this excellent set of tools alongside APP and PS

Tony

 

[Ratlet]

Really nice images

[malftobe]

 Very nice images - fantastic results!  I've also finally downloaded the trial of pixinsight.  After bookmarking the Lukomatico's set of tutorials some time back, I'll hopefully get to follow along and do some processing tonight to get a good overview of what this software can do! 

 

Quick question @mightymonoped - I'm not planning to add Blurxterminator just yet - I think I have enough on my plate getting get my head around the basics in Pixinsight.  As the tutorials on youtube from Lukomatico predate Blurxterminator, at what point did you use this process in your workflow? 

Edited February 7, 2023 by malftobe
spelling of Lukomatico

[mightymonoped]

On 07/02/2023 at 09:33, malftobe said:

 Very nice images - fantastic results!  I've also finally downloaded the trial of pixinsight.  After bookmarking the Lucamatico's set of tutorials some time back, I'll hopefully get to follow along and do some processing tonight to get a good overview of what this software can do! 

 

Quick question @mightymonoped - I'm not planning to add Blurxterminator just yet - I think I have enough on my plate getting get my head around the basics in Pixinsight.  As the tutorials on youtube from Lucamatico predate Blurxterminator, at what point did you use this process in your workflow? 

After sorting the background/colour calibration but before making image non- Linear seemed to be fine.

Tony

Edited April 29, 2023 by mightymonoped

[mrflib]

1 hour ago, malftobe said:

 Very nice images - fantastic results!  I've also finally downloaded the trial of pixinsight.  After bookmarking the Lucamatico's set of tutorials some time back, I'll hopefully get to follow along and do some processing tonight to get a good overview of what this software can do! 

 

Quick question @mightymonoped - I'm not planning to add Blurxterminator just yet - I think I have enough on my plate getting get my head around the basics in Pixinsight.  As the tutorials on youtube from Lucamatico predate Blurxterminator, at what point did you use this process in your workflow? 

BlurXterminator and NoiseXterminator are the two most ridiculous tools I have used in PI and they are both very easy to use. Compared to some of the traditional deconvolution and noise reduction tools, you will laugh out loud how much easier they are to use. Honestly I never want to open the deconvolution process again. It will make learning PixInsight easier in my opinion - though you obviously should learn the traditional non-AI tools. I am new to processing myself and am going down the same road, just a bit further ahead in PI it seems.

Star halos too big? Lower the star halo number. Stars to big? Lower the star size number. Easy.

Watching blur exterminator reveal the data that is in your image, especially nebulosity or dust, is the single most satisfying thig I have ever done in processing (other than maybe that inital STF on a lum frame!). I zoom in 1 to 1 on a section and undo-redo over and over. To be clear, it is not creating / making up data. Russ, the author of BXT, trained the AI on the highest possible resolution examples of deep space images he could. He then purposefully buggered the images up with rotation, blur etc. and ask the AI to work out how to get back to the orignial image. This is the pattern the AI is doing - identifying "problems" and getting it back to how it was if they were not there. It's a big warm safety net that can help fix bad stars, to some extent, also. It's not adding data from any other source than your scope.

Processing goes something like this:

Quote

WBPP > ChannelCombination RGB > CROP (Unless done in WPBB AUTOCOP) > Image Solver (if needed) > DBE/ABE > SPCC (colour correction, again really easy to do and free) > BlurXterminator > NoiseXterminator (both work best on linear data before any processing other than DBE) > StarXterminator (so you can process RGB without stars and add them back later).

All the above happens before a stretch. I have tested BXT and NXT on non-linear data and I was still pleased.

I prefer to remove stars as soon as possible, as above, in linear. This guide is excellent for re-combining stars: https://www.nightphotons.com/guides/star-addition . I especially love the section entitled Combining Stars with the Starless Via Relinearization (Manually).

If you are going to use Russell's tools (BXT/NXT/SXT) and have an Nvidia GPU then follow this guide to take processing time down from minutes to seconds: https://rikutalvio.blogspot.com/2023/02/pixinsight-cuda.html

In conclusion, BlurXterminator (other that PI) is the single best money I have spent on this hobby. It's like doubling my telescope aperture and doubling my exposure time. Truly incredible. A long video here but some great tidbits of info below. One of the most amazing parts of it is when Adam Block takes his data before BXT and compares it to professional telescope data. He then applies BXT to his data and compares - and BXT manages to correct his data to quite accurately represent huge Chilean telescope data!

 

[scotty38]

@mightymonoped wicked images!

[malftobe]

Thank you Tony and Peter,

 

BlurXterminator looks an awesome tool and I must admit the tool is a huge factor in giving me a push to learn Pixinsight.  Currently reading through the Blurxterminator thread but feel I'm getting a bit ahead of myself.  I have only just done my first WBPP on a small amount of data.

 

Really appreciate the details @mrflib - especially the processing workflow diagram.  Anything to make my life easier is always a good thing!

 

I only  have onboard graphics (as I found out earlier today as I had to update the drivers on my laptop for processing as Pixinsight crashed when I zoomed in on the results of WBPP).

 

I hope I am not derailing the thread @mightymonoped - I am a fair bit behind you in beginner terms!

Edited February 7, 2023 by malftobe
spelling

[scotty38]

To be honest the XT processes along with Starnet are about the limit of tools that will use a GPU so I wouldn't fret about it too much. I just built a new machine with a very fast processor and went with a cheaper AMD GPU as it wasn't worth spending double to get the equivalent performing Nvidia GPU just for a couple of PI processes, especially given I don't have Russ's tools.

The thing that really impacted me was the speed of WBPP and that is now blazingly fast as it doesn't need a GPU anyway.

 

BTW I only bought the GPU to dabble with a few games....

Edited February 7, 2023 by scotty38

[ONIKKINEN]

Looks like you got the hang of it already, not much to complain about.

Maybe one thing, looks like you NoiseXTd the starless layer a bit too much, or BlurXTd the stars a bit too much. Either way the nebulosity looks a little bit soft compared to star sizes so the nebulosity could use some sharpening, or just a reduction in image size by binning x2. Most apparent in the first image where the difference between the sharp and very tastefully done stars and the soft nebulosity is easy to see.

Edited February 7, 2023 by ONIKKINEN

[mightymonoped]

Thanks for the advice all 🙂 Looking forward to configuring and using this stuff with a bit more sophistication ongoing. Have only managed to grab a few hours in between everything else recently but will put some serious time into all of the above ongoing.

FWIW - I currently use Astro Pixel Processor for integration and cleanup (light pollution and star calibration). I'm happy with the integration side of things for now but may look more closely at the colour calibration in PI.

Hardware wise, I use an M1 Mac Mini with 16Gb and it is never noticeably slow so no complaints there.

Tony

[mrflib]

1 hour ago, scotty38 said:

To be honest the XT processes along with Starnet are about the limit of tools that will use a GPU so I wouldn't fret about it too much. I just built a new machine with a very fast processor and went with a cheaper AMD GPU as it wasn't worth spending double to get the equivalent performing Nvidia GPU just for a couple of PI processes, especially given I don't have Russ's tools.

The thing that really impacted me was the speed of WBPP and that is now blazingly fast as it doesn't need a GPU anyway.

 

BTW I only bought the GPU to dabble with a few games....

I think the main advantage of it is the ability to quicky iterate BXT and NXT incase you were not happy with the settings. This is good because BXT can sometimes affect SPCC's colour and having to re-run selecting Lum mode only can add time.

It should be noted that these can be run on preview boxes much quicker in its dedicated preview mode.

[scotty38]

15 minutes ago, mrflib said:

I think the main advantage of it is the ability to quicky iterate BXT and NXT incase you were not happy with the settings. This is good because BXT can sometimes affect SPCC's colour and having to re-run selecting Lum mode only can add time.

It should be noted that these can be run on preview boxes much quicker in its dedicated preview mode.

Totally agree.

In my case, though, I've had such a jump in general performance that these tools run quite quickly anyway - I did try the trial for BlurXT

Just with WBPP, loading it with files that would have taken my old machine in excess of 4 or 5 hours is now completing in under 15 minutes so I can now try different WBPP approaches without having to wait days on end 🙂

[malftobe]

Just a little note to add my thanks to @mightymonoped for the inspiration with the fantastic images in the original post and the other contributors on this thread. I've finished my first attempt at pixinsight processing this evening - it'll take quite a few more attempts for me to gain the confidence to post my results!

 

And another hearty recommendation for the Lukamatico pixinsight beginners tutorial videos on youtube as mentioned in the original post - really easy to follow videos that showcase some extremely powerful tools (tools with interesting sounding names for a newcomer to pixinsight that are dotted around in different menus). 

 

The videos are perfect for a beginner like me to evaluate what  Pixinsight can do for my images by guiding you through the processing.  Resulting in me being highly impressed with Pixinsight even on my first try when I have been  apprehensive about trying it in the past due to its perceived complexity.