Image processing: Then and now

[tomato]

With my latest capture stalled at 30% completion due to clouds, I was looking over some archive files and found my entry for the IKI Observatory M33 processing competition in March 2021. Back then I had no RC XT tools and I was just dipping my toe into PI, so I thought it would be fun to put that data through my current PI/RC dominated processing workflow. The result took about 20 minutes vs about 2 hrs on the original when I was wrestling with noise, colour balance and bloated stars. The latest version more closely resembles reference images I have of M33, but is a bit sparse on detail compared to the original. Although I prefer it I still can't see it winning any prizes but interestingly it does more closely resemble other recent renditions I have of this galaxy with my own data. 

Maybe the advent of AI in processing will result in less diversity of processing results, assuming it is adopted by everyone?

PI/XT tools

2021 Effort

 

 

[kirkster501]

Both are nice.  However in the version with the X-suite tools you have lost your star colour and the sky is too dark. Also some of the HII regions have gone. I actually prefer your 2021 version, personally speaking.

[ollypenrice]

I rather agree with Steve, above. M33 is a rogue case for the X-suite because, as Oskari pointed out on another thread, it is speckled with stars.

For me, the first image is very colour-cold and the second a little colour warm. I'd be no more inclined to believe PI's photometric CC than I'd believe the late Mystic Meg. The problem is not the integrity of the program but its sensitivity to input error.

Cool down the second image a little and I'd have that one any day. (I had similar misgivings in my own reprocess of M33, but it was the only one in which the new methods didn't knock the old ones out of the park.)

Olly

[tomato]

Thanks for the feedback Olly and Steve, as it happens after seeing the images together I tweaked the original and gave it a dose of NXT, somewhat on the lines of Olly's suggestion:

[Paul M]

I'm no image critic (not withstanding this critique!), I'm just an interested party. But M33 is maybe my favorite galaxy, so if only aesthetically, my view might have value

The top image is certainly M33 but as already noted, it's a bit "cold".

The "old image" is every bit M33. It's the kind of image I can look at for ages and just marvel at. It's natural and easy on the eye and beautifully presented!

The final, Hybrid, image looks too "dusty", and maybe a bit cold again?. I know, I know! Things have moved on. But I'm an old romatic.

Oh, dear, I've known it for a while but, I'm really not going to get on with many of the modern processing tools. Particularly those of the more recent AI stylee. 🙈

[tomato]

With regard to colour, the rationale behind the PI SPCC tool seemed to make sense to me so I try and use it every time. However, there are numerous occasions when it doesn’t look right so I go to CSC in APP, and when that doesn’t deliver, I adjust the channels manually to try and get close to reference images I’m using.

Why doesn’t SPCC with its 215 million star database work every time? I don’t know.

[gorann]

I think the original one is still the best. In the third version you lost the colour and brightness of the core, making it a bit flat. Keep working on it Steve👍

[tomato]

I'll keep plugging away at this until either:

a) the clouds lift and I can get my own data.

b) At least one person declares that my latest effort is better than the original, with the help of RC I am supposed to be getting better.🙂

This version used LRGBHa combination in APP and I boosted the blue channel. Then processed in PI and AP but steering clear of SXT and any colour calibration tools. 

 

[gorann]

Steve, I am that one person. The last one is the best and better than the old version in my opinion!

Unfortunately for me, I now realized that I also submitted an entry to that IKI competition, and when I look at it I can see the need for improvements, using new XTerminator tools......

[wimvb]

2 hours ago, tomato said:

with the help of RC I am supposed to be getting better.🙂

If you can get the same result in much shorter time, isn’t that an indication of improvement?

[tomato]

I agree Wim, I don't mind logging the hours to capture the data, but if I can achieve an optimum result with 30 minutes of processing as opposed to 3+ hours, I would consider that to be progress.

[gorann]

Sorry for hijacking the post, but you got me into it Steve, and here is my short process of the data using the novel XT methods.

[tomato]

That's nice, I was trying to accentuate the dust around the core, you have done a great job on that.👍

[ONIKKINEN]

10 hours ago, tomato said:

I'll keep plugging away at this until either:

a) the clouds lift and I can get my own data.

b) At least one person declares that my latest effort is better than the original, with the help of RC I am supposed to be getting better.🙂

This version used LRGBHa combination in APP and I boosted the blue channel. Then processed in PI and AP but steering clear of SXT and any colour calibration tools. 

 

Yep, this one is in my opinion an improvement to the previous ones. Perhaps a little cold but this is probably just a matter of taste.

Though here in this image we can see what Olly already mentioned before about the Xterminator being not so plug and play for M33. I think these are BlurXT artifacts where the individual stars in the spiral arms are connected into a squiggly wormy lattice type thing where they look like a single mass connected by some bridge:

Once you notice it i think it gets hard to unsee. I dont think i ended up using BlurXT for my previous M33 last autumn even when i reprocessed it a few months later (when BlurXT released) although processing is an ever continuing uphill battle so maybe today i would do it differently if i have learned anything since then. Looks like reprocessing my last M33 is on the menu next!

[gorann]

On 11/05/2023 at 23:43, ONIKKINEN said:

Yep, this one is in my opinion an improvement to the previous ones. Perhaps a little cold but this is probably just a matter of taste.

Though here in this image we can see what Olly already mentioned before about the Xterminator being not so plug and play for M33. I think these are BlurXT artifacts where the individual stars in the spiral arms are connected into a squiggly wormy lattice type thing where they look like a single mass connected by some bridge:

Once you notice it i think it gets hard to unsee. I dont think i ended up using BlurXT for my previous M33 last autumn even when i reprocessed it a few months later (when BlurXT released) although processing is an ever continuing uphill battle so maybe today i would do it differently if i have learned anything since then. Looks like reprocessing my last M33 is on the menu next!

Yes, I noticed that BlurXT in this case caused these worm-like artifacts so I gave up on using it on this data, or rather I think I added about 25% of it. But then I used the default settings. Maybe playing with the sliders would give a better result.

Edited May 13, 2023 by gorann

[tomato]

Yes, I’ll run it again with BXT turned down but…

The clouds are clearing in Shropshire!😊

[ONIKKINEN]

On 13/05/2023 at 10:12, gorann said:

Yes, I noticed that BlurXT in this case caused these worm-like artifacts so I gave up on using it on this data, or rather I think I added about 25% of it. But then I used the default settings. Maybe playing with the sliders would give a better result.

Unticking automatic PSF from the nonstellar adjustments portion and manually setting the value to significantly smaller than the actual measured PSF diameter gets rid of the worms while still sharpening the image. Testing with my own M33 i find that setting the PSF diameter to somewhere between 40-60% of the actual measured value gives a result which is tastefully sharpened (to my eyes anyway).

The issue could be with how BlurXT determines which type of deconvolution it applies, since this should be mostly stellar adjustments based but the worms come from the nonstellar portion of the tool.