Pixinsight versus Photoshop - first impressions

[Stub Mandrel]

Today I revisited some of my initial attempts at Pixinsight while I was away in Photoshop. A possible disadvantage for PI was that I processed on my laptop, whose screen though good, is rather different and fussy about the angle it is viewed.

I found some very consistent differences:

PI seemed to really flog the data hard, pulling out faint nebulosity but at the expense of poor control over stars which bloat and poor control over noise. I suspect this is why star masks are so critical in PI (I haven't mastered them yet). It may also be why many suggest noise reduction on the linear data before combining it in PI.

PS gave much more natural, less noisy results, but with less faint detail.

I felt difference was more subtle and marked than simply PI stretching more aggressively than PS/DSS, but has to do with differences in curve shape with PI (and Sharpcap's preview) using a simple curve while DSS and PS encourage a non-linear S-shaped stretch.

PS has more sophisticated control over colour balance, although PI brings out faint and subtle colour better, especially when Oiii or Sii signals are weak.

In all three images, I ended up using a blend of the PI and PS images with PI mostly contributing colour and lifting faint nebulosity, while PS mostly contributed luminosity for much tighter stars and better definition in brighter nebulosity.

 

Noise reduction in PS is easier to use, but the jury is out on which is more effective.

 

One area I found a huge difference was that GradientXterminator is not just easier to use than DBE but gave consistently much, much smoother and better results. It also did a more accurate job of balancing background colour - I actually had to use it on some of the PI images whose backgrounds were severely mottled no matter how I applied DBE before they could provide an acceptable colour layer.

It's clear I have a lot to learn about getting the best out of PI, especially about controlling stretches and noise reduction. It seems using screen transfer function and transferring it to the histogram stretch is a crude way to proceed, but good for finding what is in the image. I also need to understand star masking and reduction. DBE seems hobbled by the default approach and using bigger control areas seems to work better but even then placement is critical, and it seems small variations can introduce artefact gradients that spoil the background.

I don't think it's fair to do side by side comparisons yet, but here are the first three 'best of each' images I've done.

 

 

[Stub Mandrel]

Here's one that did seem to work well in Pixinsight, but it clearly wasn't a suitable subject for background neutralisation or gradient removal...

Instead, I just tweaked the colour balance in PS and ran a round of star reduction, then denoised in Astra Image.

 

 

[jiberjaber]

Rogelio's book, Mastering PixInsight goes to some length to underline that whilst DBE can do normalisation, it is not it's primary objective.  One thing I did find useful is less is more with DBE, I reduced the sample points from 10 to 7 and got better gradient removal as a result. I always run DBE after a suitable crop and before any channel combination.  I find the Photometric colour calibration to be excellent! If it's narrow band, then DNALinearfit script rocks prior to combining! 

The problem I've concluded with PI is it's built from a scientific perspective whereas PS is from an artistic perspective - these differing approaches suit different people I think - so whatever works I guess

PI is very garbage in = garbage out when it comes to workflow, I think PS (plugins etc) sound perhaps more forgiving by limiting complexity and choice to give an easier experience?

Whilst I've got PS, the most I've ever done in it (other than composite widefields) has been messing to clean up a mask for reuse in PI 

[MarkAR]

Had a similar situation with my Wizard Nebula where there wasn't really much room to pick background for DBE.

I'd already done a Linear Fit using Ha as my baseline and found that there weren't any real gradients to remove at all when I tried ABE, in fact it made it worse !!! So I ended up not doing either.

[Stub Mandrel]

12 hours ago, jiberjaber said:

PI is very garbage in = garbage out when it comes to workflow, I think PS (plugins etc) sound perhaps more forgiving by limiting complexity and choice to give an easier experience?

I suspect much of the complexity of PI is in the interface rather than the actual processes; the choices are vast but the number that are actually effective is pretty small.

PS can do hideously complex things, there are actions (not written by me) that use many dozens of steps to achieve great results.

Both programs have a fundamental weakness in offering you curves to manipulate that are tiny, forcing you to resort to entering or adjusting numbers. The best thing about PhotoPaint is that its graphs are approximately twice the size.

[Ouroboros]

For curves I click the curve where I want to adjust it and then use the up/down/left/right buttons for fine control. 

[Ouroboros]

I agree the interface is a bit overwhelming. I found watching some of the introductory video tutorials on YouTube very good to see how others use the interface. It’s actually very logical once you see it used. I particularly liked Astrodude’s videos  in this respect. He creates instances of the processes he’s going to use and arranges them in order of his work flow down the right side of the screen. He then saves that set of processes for recall later. There are probably other ways to work but that made sense to me. 

[Ouroboros]

Something I’ve never got my head around is saving work as a Project In PI.  So what I do is save each stage as I apply each process so that I can backtrack to any stage if I want to. I used to do that in PS too. 

[dannybgoode]

4 hours ago, Ouroboros said:

Something I’ve never got my head around is saving work as a Project In PI.  So what I do is save each stage as I apply each process so that I can backtrack to any stage if I want to. I used to do that in PS too. 

Yes this is what I do also. I have a base ‘project’ which I load up and that has all my basic processes set up already so for the initial work I just go through them in turn but save the outputted subs each step. 
 

I am now pretty comfortable with the basics but need to explore the more advanced processes in much more depth. Wouldn’t be without PI though-I can certainly do much more with the raw image data through to integration than I can using any other software I’ve tried. 

[Ollyb]

I’ve just started using Pixinsight after seeing the results other propel have achieved, it’s a whole different concept to Photoshop and seems to be far less destructive of faint data…

Here is my best attempt at NGC7000 in Photoshop and then in Pixinsight… They are totally complementary to each other but I’m loving the results from Pixinsight, especially the ability to see faint detail…

 

[johnturley]

I currently have Adobe Lightroom (not the full Photoshop package), which I purchased at a bargain price.

I have however considered getting PixInsight, but am put off by the complexity, and the high price compared to Lightroom, I gather it costs around 230 Euros to purchase and download PixInsight, although I gather that you can get a free trial version.

I addition I probably wouldn't use it enough to justify the high cost, and furthermore I am more into lunar and planetary imaging (I have recently purchased a ZWO ASI 462 Planetary camera) than deep sky objects,  I gather the advantages of PixInsight would be more apparent with the latter. 

I also find it annoying that you can no longer purchase most software on CD/DVD and instead have to download it, which sometimes can be tricky trying to get the right program to download and install (often multiple websites appear), plus you are usually limited to installing it on just one PC.  

John 

[david_taurus83]

9 hours ago, johnturley said:

 

I also find it annoying that you can no longer purchase most software on CD/DVD and instead have to download it, which sometimes can be tricky trying to get the right program to download and install (often multiple websites appear), plus you are usually limited to installing it on just one PC.  

John 

Pixinsight can only be downloaded from the developers website. It can be installed on several PC's at any one time with a single licence. I currently have it on 2 laptops and even on my scope mounted mini pc.

[wimvb]

On 18/08/2020 at 21:37, Stub Mandrel said:

It seems using screen transfer function and transferring it to the histogram stretch is a crude way to proceed

I never ever use stf as a permanent stretch. It's way too aggressive. I most often start with histogram tansformation (levels in ps), with the midpoint shifted to 0.25. I apply that several times, with the black point brought in slightly if necessary. When the histogram top is clearly separated from the left side (at a position between 0.15 and 0.2), I switch to curves transformation, bringing the histogram top down to about 0.8 and lifting the mid range to get an S curve. I try to keep the upper part of that S curve as straight as possible, to avoid excessive star bloat.

PI allows you to zoom in on any open image window, including the graphs in histogram transform, curves, color saturation, etc. You can also resize these windows to your liking.

When I do deconvolution, I adjust the stf stretch, making the nebula/galaxy a bit dimmer, so I can see details in the brighter parts. Just open the stf tool, apply stf (not as a permanent stretch) and check the check mark at lower right. Then zoom in on the sliders for black point and mid point. These are usually almost on top of each other, but when zoomed in , you can control them better. Pull the mid point slider to the right and the black point slider to the left, in order to soften the screen stretch.

 I've made the comparison before, but the difference between PI and PS 😉

PI:

PS:

[scotty38]

11 hours ago, david_taurus83 said:

Pixinsight can only be downloaded from the developers website. It can be installed on several PC's at any one time with a single licence. I currently have it on 2 laptops and even on my scope mounted mini pc.

I just paid up for PI after trialling it a couple times and I'm going to put Linux back on one of my machines to give it a go there too given it's their preferred platform. I'm a massive novice but what swung it for me was the masses of quality videos to help in learning it. I've been particularly taken by Adam Block's training and like his style.

BTW the "available training swung it" comment was versus things like APP for example not Photoshop as I realise that has plenty......

Edited June 4, 2021 by scotty38

[Clarkey]

On 04/06/2021 at 08:53, wimvb said:

 I've made the comparison before, but the difference between PI and PS 

Perhaps a little harsh. However, it does also highlight the one major failing in PI - it is completely daunting to newcomers. I could work the second one - the first would just cause me to run away🤣

I suspect it is the best option overall. However, I use APP, Startools and Affinity because I actually understand what I am doing. At some point I will consider PI again, but when I watched one of the training videos I still did not really understand what they we doing!

[wimvb]

3 hours ago, Clarkey said:

Perhaps a little harsh.

Perhaps. But at the same time, the comparison isn't completely off. Both instruments allow you to create great sound, but the larger one gives you more control over every aspect of the process. This is daunting at first, and the learning phase will be longer. But once you learn to use all the controls, you can achieve great results.

One valid point against pixinsight used to be the lack of documentation. But imo, this point has lost its validity.

I think you should try again. 🙂

[City9Town0]

When I was solely using PS I kept revisiting data as I gained new skills and this produced improvements.

I now have PI and keep revisting data as I gain new skills and seem to get even greater improvements. There do also seem to be more and more 'training' resources for PI.

If you can surmount the myriad of challenges getting good data in the first place then you may well have the application to enjoy the challenges and subtleties of PI.

If finances allow, PI+PS enables you to make the most of both.

Daunting... so many things appear daunting at first; for PI I used the Light Vortex tutorials and within the trial period had exceeded my PS results.

Could I survive without PI... YES. Do I enjoy time spent using it... YES. In my case I believe that my results are better with it.

Other opinions on PI are readily available!

 

[Clarkey]

Maybe come the darker nights I'll give it another go. I appreciate that it probably is the best tool for the job once learnt- but unfortunately time is a very precious resource for me. (7 and 9 year old boys and full time job). However, now I have largely got to grips with the data gathering I guess this is the logical next step.

[AbsolutelyN]

I don't think it's an either or. Both are incredibly powerful image editing software packages.  You can do everything in either of them but personally being already very familiar with photoshop I use a mixture of both and I find they compliment each other very well.  PixInsight offers very specialised tools which makes many things a lot easier - gradient removal and star reduction for example. So if I was to pick one for astro I'd go with PixInsight, but both make for a very powerful combination.