›

Content request; how to process 'normal quality' images

Helge Kooistra
in PixInsight
I appreciate all the content Adam produces, but for me (and maybe others) there is one drawback: 

Most images used are inheritly of very good quality. They come from telescope live, semi-professional observatories or people with mega telescopes.
But i (and likely a lot of people here) have more modest equipment what results in a lower quality images to begin with. I image from a Borlte 9 backyard with a triplet refractor 115/800 with a .8 reducer, and a ZWO ASi294mm Pro camera. Gives me lower resolution to begin with.

 
I struglle to follow the tecniques used because sometimes they result in different outcomes for me. for example: when Adam makes the step to combine the superluminace with the LRGB it results for him in a detailed image, for me it gives the image a boost but not so detailed, looks more 'feathered'. I think because the detail is just not there.
'
I understand its maybe not so much eye-candy for the course, and you definitely won't win an APOD with the end result. But for us modest equipment owners who try to do their best it can be of great help to see whats possible with our kind of image quality.

Comments

  • BlockHeadBlockHead
    I did have examples that start from data that is perhaps more challenging because of light pollution, lack of exposure time and so on. 

    Here is the thing, members wrote back to me. They said that they did not want to see all of the extra information and specific case-by-case steps. They said it was confusing to them. I think part of the reason this is true is because people do not like to watch me solve problems that they do not have in their own data. 

    For example, I have an example of M51 where the person's Red data was a little out-of-focus. So I showed a technique that still allowed the incorporation of this information. But I heard back from people that they felt this was a side-track and distracting. They just wanted to see the flow.

    In another example I processed some images from a member's data that was from Detroit (Bortle 20).
    Similar reactions.

    So the techniques to follow are more principles then specific steps. Adding "superluminance" isn't supposed to add detail- it should have a bit better S/N (a little less noisy). That is supposed to make it easier to process than a slightly noisier straight RGB combination. It does depend on the data if this has benefit. 

    I think it would be better to look at some cases where there is data like yours- and we look specifically at where there is an issue (like the L inclusion). So perhaps you can organize your question with some screenshots and show the issue. Then I can perhaps make some suggestions- or as you advise, make a video that highlights some specific and hopefully common issue. It could be that I simply point out- hey look at this kind of data... adding a superluminance will not help because of a,b,c ... 

    -the Blockhead
  • Helge Kooistra
    I understand the reasoning, thanks for clarifying! 

    My end result:
    To mee its all a bit overcooked, maybe i should be more cautious with processing (or accept i cannot pull out of the data what's not in)

    When zoomed in i hope it's visible what i mean by 'feather like' appearance
  • Helge Kooistra
    And to show where it did go downhill (All steps show the RGB and the Superluminace at the same step -and i know the superluminace is not meant to bring out extra detail, i meant the Luminance- ) 

    Before BlurXT it was ok

    After BlurXT seemed ok to me

    After noisext its still bearable


  • Helge Kooistra
    My masters (if you are curious...)


    A total of 30 hours from Borlte 9 skies (Sky quality of about 18), 115/800mm triplet refractor with 0.79 reducer, asi294MM pro camera. Pixel scale of the system 1,5 arc second
  • Helge Kooistra
    And i appreciate the reaching out! 

    There's a huge gap in knowledge between where you stand and my position  ;)

    You can think ahead and see what little step has to be made first, and you seem to think in techniques. I'm still in the functionality-phase like  'i want to improve the contrast a little bit' or 'that part needs sharpening' and then get lost in processes and scripts. So i'm re-playing your videos and trying to make notes to get the kind of knowlegde base i need (improving color? these processes/ scripts with the settings that can be tweaked. Bringing out more faint details? these processes, etc)
  • BlockHeadBlockHead
    I get it. 
    You might not like this answer though- 
    The reason I seem to know more isn't because of some magic stuff or skill... it just comes from practice. 
    Now, when I practice (process different images over and over) I do pay attention to patterns. Things that
    work and things that do not. I look at colors, and numbers and settings and stuff. 

    But attentive practice is really the key.
    That being said, my demonstrations I hope offer insight into possible ways to proceed with your own data.

    I will try to look at your masters when I get a chance...
    -the Blockhead

  • BlockHeadBlockHead
    The Luminance looks good to me.
    So... please make links to some screenshots that show where you are ...and what you think needs to be "improved."

    -the Blockhead
  • Helge Kooistra
    I think i did just that in the post before the masters
  • BlockHeadBlockHead
    Got it... sorry.
    So... this isn't about poor quality data... not at all.
    This is all about choices. 
    I agree with both parts of  "To me its all a bit overcooked, maybe i should be more cautious with processing (or accept i cannot pull out of the data what's not in)"

    It *looks* like to me you process data when you are NOT looking at the image at 100% or more.
    People tend to process the image to look good when zoomed out. This is wrong. You can tell me you did not do it...I am rarely right...but if you look at the zoomed in version of the image I do not see how you got there without noticing things might be going wayward. Processing should always be done in very small increments. For contrast enhancements...doing something twice at 50% is usually better than once at 100% of the effect.

    I suspect you applied NXT way too early. You generally want to do this at fuller force near the end of processing on a stretched/enhanced image. Doing a little early is OK...but only a tiny amount. 

    You should probably indicate some other things you did.

    -the Blockhead


  • Helge Kooistra
    edited April 29
    Its a shame my data quality is ok, i was already eying an Edge HD8 but you just ruined my businesscase  ;)

    First: no i don't process at 100%, so you're absolutely right about that. I zoom in, but also regularly take a step back and do things that look good zoomed out. So i'll definitely take that advice!

    I Shoot with a camera that natively is bin2 and had 4.63 microns per pixel but its also possible to unlock bin1 mode with 2.315 microns per pixel. But that gives you craze big images so i shoot in bin 2 and then use 'drizzle 2' with drop shrink of 90 at post processing in WBPP. Don't remenber exacty why but it gives me roudner stars. Maybe i should scale back the image before processing to get more detail to begin with. 

    The steps i took
    - combine RGB
    - starxt
    - DBE on starless
    - put stars back in
    - spectometric color calibration
    - SCNR
    - BlurXT
    - Noisext 
    - Starxt
    - stretch image
    - colorsaturation
    - LRGB combination (add superLUM to RGB)
    - Localhistogramequalisation (kernel 22, contrastlim 1.6, amount 0.36)
    - Localhistogramequalisation (kernel 74, contrastlim 1.6, amount 0.36)
    - Curves (slightly dim rgb and R)
    - add stars
    - Curves (boost colors, add more contrast)

    I looked back in the process explorer and redid all the steps zoomed in at 100%. There are some very small galaxies in the lower left with a kind of shockwave. Looks like things went downhill after the first LRGB combination where the contrast seems to disappear a bit


    But the final blow was the second blurxt that made it looke more like a watercolor painting with too much paint used.


    And to top it off the last curves 'enhancements' made the it much worse.

    I should have left the image more soft, not try to let the colours pop that much

  • BlockHeadBlockHead
    A couple of comments...

    1. BXT early on needs to be done with care. Since you are doing LRGB...it is more important to do it to the RGB component..but not the L early. 
    2. BXT can now be used to operate on high or low frequency as well as luminance or chrominance... you can use this to your advantage instead of just hitting with a big hammer (which it seems me to you did).
    3. The use of LHE is one to do with care- You might consider blending only important elements of the result (to protect other areas). Also, you know I show a method of applying this with conditions or bright or dark features being enchanced to different degrees. You might want, for example, to affect the dust lane more strongly in an image like this than the brighter HII regions and things. These are darker features..and you can have LHE and other tools operate more on this rather than both bright and dark simultaneously. Did you consider this?
    4. Yes, I do think that BXT twice in combination with applying it zoomed out is the issue. 

    -the Blockhead
  • Helge Kooistra
    I tried a new -very more careful- approach and have to say it worked out OK for my liking...

    There may be a lot that can be improved, but i think i did the 30 hours of exposure more justice. 
    - the image is much softer, background is not pitch black and the distant galaxies are soft but visible. Let's hope i didn't wipe out entire civilisations this time...
    - details are visible but not forced
    - colors are within limits, for like likings the HA pockets are too much but i like it
    - Fainter details are visible, without completely blowing out the core

    https://drive.google.com/file/d/13yxFyLWAibsAG97xzhDNwK1mYS-7qrK4/view?usp=drive_link

    The main difference in approach: 
    * Processed everything more zoomed in, and in very small steps
    * Created a superLUM, combined the LRGB directly, first steps of processing and stretching and then separated LRGB and worked on the Lum. This way i was certain i stretched the Lum the same way. 
    * in the end i rescaled the image back to its original state after the drizzle 2x during WBPP to hopefully show more detail...

    Thanks for the advice!
  • BlockHeadBlockHead
    Let me check it out...
    (you need to change permission to anyone that has the link..)
    -the BLockhead
  • Helge Kooistra
    Sorry, I changed the permissions, now it should work
Sign In or Register to comment.