1.8 closed alpha feedback

General discussion about StarTools.
Mike in Rancho
Posts: 1141
Joined: Sun Jun 20, 2021 10:05 pm
Location: Alta Loma, CA

Re: 1.8 closed alpha feedback

Post by Mike in Rancho »

Good stuff gang, I will have to try a few of those things. :thumbsup:

I'm getting better, but not there just yet. A few more observations, again just on my M5 data (I'll have to start trying other data too):

I'm getting a few anomalies in auto star masks - little black dots. Well, black if autogenerated in say, Sharpen, but green if you just manually go into mask for an auto-stars-do. I pixel zoomed and there is nothing there, so not sure what's triggering it. I restored and tried without Correlation in Wipe, but same result so that's not it. Not a huge deal.

I like the correlation filter - would love to know more about how to objectively/subjectively select the level. It does change how you AutoDev. Some likely due to the ignore fine detail level, but seems beyond that too. So you have to alter your ROI a bit from one used without the correlation filter.

I still get no effect at all for the "dark side" in Sharpen.

In synthetic deconv, the mask doesn't seem to do anything. I tried with the star mask (really apod mask), full mask, and no mask.

Blue box selections are not saved along with the mask/apod mask. Which I kind of expected. So, sampling from scratch every time.

Deconv in 1.8 seems to need more retained resolution than in 1.7. My usual crop and 71% bin on this data leaves me at about 2300 x 1500, roughly, and the little stars (pretty much all a glob is) mostly just end up heavily squared. I will try again without the bin (I wanted that SNR lol) and see how it goes.

My new method of gray hole repair is this - select just the gray hole (similar brightness works well) pixels, and go in to Layer. Change the mode to Add. This should saturate out the gray hole back to white. If that is then too harsh, hit grow mask once or twice and go back into Layer, and click up the kernel radius for a blur. Really even once click (1.1) should do the trick. Of course right now I am just seeing the one gray hole on the one big star in my image. Lots of them could be a bit more laborious.
prusling
Posts: 23
Joined: Tue Feb 06, 2018 2:21 pm

Re: 1.8 closed alpha feedback

Post by prusling »

hixx wrote: Sat Jun 26, 2021 6:12 am Hi,
There is a way to drive COMPOSE module into crashing....
Just load your LUM file and while loading wheel is still turning, hit the RED button.
GUI comes up with the directory, try load quickly. Now 2 loading tasks seem to be in the queue which brings whole Star Tools to a grinding halt...
No way to get out except quit. (process works with different colours too)
Probably ST should either enable parallel loading jobs or prevent users from starting loading jobs in parallel or just put them in a queue and execute these jobs one after the other, independing from "nasty" user behaviour :-)
This bug was present in 503 and has not been resolved yet, not sure if this is already known

reagards,
jochen
Hi, I have experienced this issue on 1.8 and previous builds - if you don't allow a load to fully complete, and hit another load button too soon, ST can hang. I try and discipline myself to wait but still get caught occasionally!
Carles
Posts: 191
Joined: Thu Jul 30, 2020 5:06 pm

Re: 1.8 closed alpha feedback

Post by Carles »

Hi,
Testing ~504.

I have a question though...
NBAccent file, gets Wiped and "Stretched" as the other files (let's assume Luminance and RGB in Compose) but it seems it doesn't get either Contrast, HDR , Sharp and Deconvoluted.

I was playing with the the M27 files I have with Lpro and L-eNhance, and when I got to NBAccent, the image got "soft" , it seemed that all the aforementioned modules work-tweaks got "layered up" by the L-enhance file... Hence I wonder what processes affect the NB file. After the use of the module went to HDR again, but can't get same effect (as my NB file has stronger luminance core, and if I use ROI in the previous step to get that HDR effect in the M27 core, then I loose the outer details. )

Will try to reproduce the workflow later.
FYI, I have tested with parameters in NBAccent module too.

So in that case I was wondering if, in the case the NB file doesn't get those detail enhancement modules , how would it affect to process the NB file with buton "open" , and then use that as NBAccent file in Compose ( althought, NB file would not be linear anymore...but does that actually matter after decon and so on has been done to the "main" file? ")


Regards

Carles.
hixx
Posts: 250
Joined: Mon Sep 02, 2019 3:36 pm

Re: 1.8 closed alpha feedback

Post by hixx »

Hi,
just processed an RGB stack (loaded using OPEN/linear w/out Whitebalance), and found the NBAccent module (not greyed out), even without any narrowband dataset loaded. Launched it, just curious what it would do. Screen 1 appears with a mutilated version of the image - looks like it loaded the R channel only into the narrowband slot. It this a bug or a feature?
jochen
hixx
Posts: 250
Joined: Mon Sep 02, 2019 3:36 pm

Re: 1.8 closed alpha feedback

Post by hixx »

Sorry, NBA crash
Problem: too fast of a user :-)
when in second screen, click back and before first screen is rendered completely, click Cancel and its gone :cry:
excerpt of the report:

Process: StarTools [1480]
Path: /Applications/StarTools Folder/*/StarTools.app/Contents/MacOS/StarTools
Identifier: com.SiliconFields.StarTools
Version: 1.7
Code Type: X86-64 (Native)
Parent Process: ??? [1474]
Responsible: StarTools [1480]
User ID: 501

Date/Time: 2021-06-30 19:52:19.183 +0200
OS Version: Mac OS X 10.14.6 (18G95)
Report Version: 12
Anonymous UUID: 90955F03-E920-13D9-BFB7-82D6B639895B

Sleep/Wake UUID: B6E675FB-5516-4679-956D-90191AF008DB

Time Awake Since Boot: 34000 seconds
Time Since Wake: 7300 seconds

System Integrity Protection: enabled

Crashed Thread: 1

Exception Type: EXC_BAD_ACCESS (SIGSEGV)
Exception Codes: KERN_INVALID_ADDRESS at 0x0000000000000034
Exception Note: EXC_CORPSE_NOTIFY

Termination Signal: Segmentation fault: 11
Termination Reason: Namespace SIGNAL, Code 0xb
Terminating Process: exc handler [1480]


let me know if the whole report was required
BTW, report says V1.7 but it was 1.8.504....
regards,
jochen
User avatar
admin
Site Admin
Posts: 3367
Joined: Thu Dec 02, 2010 10:51 pm
Location: Melbourne
Contact:

Re: 1.8 closed alpha feedback

Post by admin »

Many thanks for your reports everyone!

I just uploaded a new version (see email you should have received) that addresses some of the issues reported.

CHANGELOG;
  • First cut of improved Sharp module, with improved algorithm, new Protection parameter and enhanced de-ringing behaviour
  • Added "Add 1/2 Unity + Clip" Cap mode to Layer module
  • Added "Gaussian High-pass filter to" Filter Type to Layer module
  • Fixed small bug in NBAccent module that would incorrectly render over-exposing areas (e.g. star cores) ("gray holes")
  • Fixed samples temporarily disappearing in SVDecon module
  • Fixed handling of dynamic range extension in SVDecon module ("gray holes")
  • Disabled NBAccent module if no accents loaded in Compose module
Ivo Jager
StarTools creator and astronomy enthusiast
Carles
Posts: 191
Joined: Thu Jul 30, 2020 5:06 pm

Re: 1.8 closed alpha feedback

Post by Carles »

Thanks Ivo!!

I was testing a file with NB , and when in Wipe module, visualising the files, only "Color" is in actual color,
NarrowBand file is BW, as well as Luminance(which is right I assume)
is this correct ?

Carles
hixx
Posts: 250
Joined: Mon Sep 02, 2019 3:36 pm

Re: 1.8 closed alpha feedback

Post by hixx »

Hi
This toggle button in AutoDev seems to be misbehaving. Currently, if no NBA file is loaded it toggles Luminance / Color / Narrowband (while Luminance file seems to be presented as NB), at which point it stops to provide any further toggle option.

It probably requires a dependency on whether a NBA file is actually loaded in COMPOSE or not.
If NBA Loaded: Luminance / Color /Narrowband / Luminance...
I no NBA file loaded: Luminance / Color / Luminance....

jochen
hixx
Posts: 250
Joined: Mon Sep 02, 2019 3:36 pm

Re: 1.8 closed alpha feedback

Post by hixx »

Theory of gray holes....
I can confirm with 505 the gray holes turned into "less bright white" holes surrounded by an overdriven ring which is not so much of an issue as before.
Nevertheless I wondered at which step in the processing chain this effect might be introduced. I found this slight darkening of the star core already present in COMPOSE, after loading Luminance file only, no processing involved at all! Hence SV Decon seems to no longer contribute to this effect. So I wonder if this might be a display / driver /graphics card issue. Can anybody else confirm this effect?

Actually the "less bright white" holes are not so bad after all , as they will restore the core color again. To me its more the white ring around the core which is a problem now. Provided this effect to be confirmed, I wonder if there was any way to introduce a special kind of 'Tighten' mode in Shrink or Repair which drags the CORE color & luminance outwards into the ring in order to restore a perfect star....
But probably it would already work to review luminance loading in COMPOSE and fix the white rings here or in Autodev.
clear skies
jochen
Mike in Rancho
Posts: 1141
Joined: Sun Jun 20, 2021 10:05 pm
Location: Alta Loma, CA

Re: 1.8 closed alpha feedback

Post by Mike in Rancho »

My guess is these are star attributes that exist at the outset, meaning in the data. It tends to be the brighter stars, so the core could be saturated. I didn't have an L file to compose, but opening an OSC file center option creates an L for "compose" mode anyway. Even the linear file, before stretching, shows some concentric differences on my 5 Serpentis. For sure the core of that star is blown out, the other rings - maybe optics? I've seen this quite a bit and not just in my data.

Things then change around somewhat depending on your AutoDev and ignore fine. HDR can also work such stars over a bit. Can't say the new SVD has no effect though.

I couldn't check for color after SVD though. My laptop is acquiring data, and my desktop always crashes at the outset of the color module in GPU, no matter what. Too bad, but it's an old machine I souped up with a Q9550 and a pretty decent graphics card, mostly just used as my stacking machine.

A lot of these ringing star defects though do need a layer blur with a custom star mask after denoise, so that's something I generally do now. It kind of has that effect of bringing outer color in while taming the center or a bright donut ring. Sounds like your wanting pretty much the opposite effect, but I imagine the layer blur would work both ways.
Post Reply