PixInsight Forum

PixInsight => Image Processing Challenges => Topic started by: Herbert_W on 2012 April 18 04:06:35

Title: vdb142 - narrowband - have a try
Post by: Herbert_W on 2012 April 18 04:06:35
Hello!

I did vdb142 with narrowband (ha 2.5h, oiii 3h, sii 5h) and it is not
possible to me, to create a pretty colorimage with the hubble palette (s-h-o).
Usually the have a light yellow and blue - but my tries ends with flat colors without blue and yellow.

On the one side is the mixture of the channels with pixelmath important (there a a lot of recommendations in the forum),
otherside this can't be all - I suppose that curves, histo and saturation important as well - but I cant get it.

Maybe there is someone, who want to try my frames and show me his result.

The files are at: endor.pixinsight.org   Folder: HerbertWalter

or

vdb142 sii (http://www.skypixels.at/downloads/PI_Forum/SII-Baader_cropped.tif)
vdb142 oiii (http://www.skypixels.at/downloads/PI_Forum/OIII-8.5nm-Baader_cropped.tif)
vdb142 ha (http://www.skypixels.at/downloads/PI_Forum/HA-7nm-Baader_cropped.tif)

Thank you and best regards.

Herbert, Austria


Title: Re: vdb142 - narrowband - have a try
Post by: Lex on 2012 April 20 04:04:39
Hy Herbert,

I have downloaded the files and will try to give it a try during the WE...
Will keep you in the loop..
Thanks

Title: Re: vdb142 - narrowband - have a try
Post by: Herbert_W on 2012 April 22 00:02:20
Hi Lex!

Thanks - I'm very interested on your result!
Its not so easy to do it in PI - its a pity to miss a tool in PI to mix narrowband frames.

Best regards.
Herbert, Austria
Title: Re: vdb142 - narrowband - have a try
Post by: vicent_peris on 2012 April 22 00:42:51
Hi Herbert,

I have a result, will show it on next week, be patient. :P


Regards,
Vicent.
Title: Re: vdb142 - narrowband - have a try
Post by: Raymond on 2012 April 23 12:53:00
Hi Herbert
I've actually not a lot of experience in combining narrowband images, but I wanted to see what I could get with your data. So here is my result...
--------------------------------------------
 1. Star Alignment
 2. PixelMath   R: Ha
              G: SII
             B: OIII
 3. Dynamic Crop
 4. BN using a small Preview
 5. CC with structure detection
 6. Masked stretch  50 iterations
 7. HT  shadows:  0.042
      midtones: 0.260
 8a. Star Mask
   Scale: 5   Growth: 1
   Smoothn: 8
 8b. HT on Star Mask
   Shadows:  0.250
   midtones: 0.420
 8c. Invert Mask

 9. ATWT with Star Mask applied
   Bias: +0.010/+0.020/+0.025
         +0.020/+0.025/+0.030
        for layers: 3...8
 10. CT   RGB/K (S-curve; contrast)
   I: 0.11  O: 0.09
   I: 0.60  O: 0.76

--------------------------------------------
Is this a result similar to yours?
Title: Re: vdb142 - narrowband - have a try
Post by: Herbert_W on 2012 April 23 21:04:41
Hi Raymond!

It looks great!! Thank you very much for processing.

Quote
Is this a result similar to yours?

If I use H-O-O for R-G-B, it looks similar to your result - nearly a "natural look" like RGB.

Very interesting is your workflow - thanks for the effort to list the detailed steps.
(although Star Alignment wouldn't have been necessary, because the frames are already aligned).

In the meantime I got 2 examples via PN. At the end I will show the results to compare
and discuss.

Thanks again and best regards.
Herbert, Austria
Title: Re: vdb142 - narrowband - have a try
Post by: Lex on 2012 April 24 13:32:25
Hello Herbert;

This is the third time I try to process a NB image...

Here is what I've done:

01. Star Alignment
02. Dynamic Crop
03. Fast Rotation
04. PM: 0.8*SII + 0.2*Ha / 0.2*Ha + 0.8*OIII / OIII
05. Background Neutralization
06. Color Calibration
07. Masked Histogram ( Starmask inverted applied )
08. DBE
09. ACDNR
10. SCNR --> Green Average
11. Curves --> Lightness / Saturation

__________________________________________

That's it; my rendition of your Data

I hope you like it; I tried to apply HDRMT but as I am kind of short in time and still struggling with the different layer setting I satisfied myself with the end result.






Title: Re: vdb142 - narrowband - have a try
Post by: Herbert_W on 2012 April 25 03:39:02
Hi Lex!

Quote
I hope you like it;

Yes, I like it - good looking colors!!
Thank you for processing and showing the result.
Very interesing are the factors to mix the channels " 0.8*SII + 0.2*Ha / 0.2*Ha + 0.8*OIII / OIII".

Did you guess your formula and hence it is a 'try and error game' or is it
a proven mixture?

At the moment I'm looking for a formula which takes the different signal strength into account.
But there is a lot to think about to measure the values, background, signal, bring it in relation, ...  and this takes time.
I'm not a mathematicien or physicist :-(

Thank you also for the detailed workflow ( staralignment would not have been necessary, its already done).

Best regards.
Herbert, Austria
Title: Re: vdb142 - narrowband - have a try
Post by: TeeJay on 2012 April 25 03:47:49
Hello Herbert,

I also tried to process your Data in classical Hubble Palette. I hope you like the result. I used the Tone mapping Method from J-P Metsavainio, with a lot of modifications from me...

CS
TeeJay
Title: Re: vdb142 - narrowband - have a try
Post by: Lex on 2012 April 25 08:32:01
Hi again,

You are welcome, in fact the channel mix factors came out of a tutorial that Juan has written.
I am no mathematics guy neither so I took the easy way, unfortunately, because i would like to have more time for trial and error what is not possible for the moment..

@TJ

I like your rendition very much; I think it is "closer" to the Hubble palette than mine. What are your Pixelmath channel factors if I may ask?

Thanks
Title: Re: vdb142 - narrowband - have a try
Post by: TeeJay on 2012 April 25 10:01:20
Hi Lex,

actually i don't use any factors at all. My first step is to HistogrammTransform all the three images to the same level. To do so, I perform "Auto zero shadows" and check it with the 100x zoom, after that I set de Midtones to have the hill at 3/16... (I attached a screenshot for better understanding)
Those images don't have the high contrast of auto-SFT Images but this is not needed.
Then I use a CurvatureTransforamtion to set nearlly the same (visuall) contrast to all the images. (screenshot)
Now I finally map [SII] to red, H-alpha to green and [OIII] to Blue, without any factors.

I know this is not a scientific way, but it is very good for pretty pictures!

In the case I showed above, i do all this with Tonemapps (starless) and use a luminance ((lum from tonemapps) + Ha + OIII + SII) at the end...

CS
TeeJay
Title: Re: vdb142 - narrowband - have a try
Post by: Lex on 2012 April 25 12:14:35
TeeJay,

indeed you use a very simple workflow. The only negative thing IMHO is that the star coloring is a bit towards red.
Have you neutralized the background and color calibrated your image?
This is really the only "negative" aspect of your rendering; again IMO :)

Title: Re: vdb142 - narrowband - have a try
Post by: TeeJay on 2012 April 25 14:01:33
Lex,

yes the thing with the stars... Because there are no usefull color information from Stars in Narrowband images, I do a simple trick: Starmask -> CurvatureTransformation -> Saturation and C to zero.
That's it. :P
No BackgroundNeutralziation needed, because of the HistogrammTransformation to the same levels. ;) But if you do to much CurvatureTransformation you will need it...

If I map narrowband images to "real color" with somethin like [R=(SII*2+HA)/2;G=OIII*0.5;B=(OIII+HA*0.333)/2], I use singel RGB exposures of the Stars to overlay.

CS
TeeJay

Edit: I reworked the Image, with better Star-removal and also did a real color version (unfortunatelly forgot the sharpening)...
Title: Re: vdb142 - narrowband - have a try
Post by: Lex on 2012 April 25 14:22:14
Why doing it the difficult way if it is possible to do it easily  O0
Nice approach  >:D
Thanks
Title: Re: vdb142 - narrowband - have a try
Post by: Herbert_W on 2012 April 26 01:28:16
Hi TeeJay,

wonderful pictures and thanks a lot to process the frames and share your result and especially your method with the community!

Your completely different approach corresponds to my idea to create reproducible steps processing narrowband images.
I spend my time at the moment to retrace your steps and I have to say, it's not so easy.

Now I'm for one week on duty travel and will be back afterwards with my 'calculated' result -  my laptop with PI is in the package ;-)

Until then thanks to all for the contributions - maybe we will find a way to leave the 'try and error' method.

Best regards.

Herbert, Austria

@Vincent - I'm very curious to your result and method ;-)




Title: Re: vdb142 - narrowband - have a try
Post by: TeeJay on 2012 April 26 03:46:27
Hi Herbert,

I can explain you the steps if you like. But i'd like to do this in Skype, as I am from Switzerland we can speak german. ;)
You can write me at astro(├Ąt)teejay.ch

CS
TeeJay
Title: Re: vdb142 - narrowband - have a try
Post by: Lex on 2012 April 26 08:09:27
...Vicent not Vincent  >:D
Title: Re: vdb142 - narrowband - have a try
Post by: TeeJay on 2012 April 26 09:31:45
I published the same workflow on another image here: link (http://pixinsight.com/forum/index.php?topic=3944.msg29288#msg29288).

CS
TeeJay
Title: Re: vdb142 - narrowband - have a try
Post by: Herbert_W on 2012 April 26 10:19:58
Quote
...Vicent not Vincent  >:D

uups - sorry!

@TeeJay!
Quote
I published the same workflow on another image here: link.



Thank you for this additionel info - now I have to study a lot ;-)

Best regards.
Herbert, Austria
Title: Re: vdb142 - narrowband - have a try
Post by: astropixel on 2012 May 03 04:34:30
Well ??? First attempt at narrow band :o

Registered and ChannelCombination SII, HA, 0III. Colour calibration, HT luminance mask, HMT, LRGBCombination L*, ColourSaturation, HMT, ATWT, LocalHistogramEqualisation, Deconvolotion, CurvesTransformation.
Title: Re: vdb142 - narrowband - have a try
Post by: Lex on 2012 May 03 06:02:50
Quote
Well ??? First attempt at narrow band :o

Nice attempt I think!

Congrats!
Title: Re: vdb142 - narrowband - have a try
Post by: astropixel on 2012 May 03 17:37:50
2nd try. Is this approaching something like it. Hubble pallet ??? I apologize in advance.

Registration
ColourCalibration - not sure if this is necessary?
Stretch, Multiscale transform, saturate, curves and noise reduction

Combined as follows;

0.8*SII + 0.2*Ha
0.2*Ha + 0.8*OIII
OIII

Thanks Juan - http://pixinsight.com/forum/index.php?topic=3734.msg25646#msg25646

Title: Re: vdb142 - narrowband - have a try
Post by: astropixel on 2012 May 18 02:33:54
Had another go at this. I think this is the correct result. I'm not sure where I went wrong with the previous image.
Title: Re: vdb142 - narrowband - have a try
Post by: Lex on 2012 May 19 14:35:35
Hi Astropixel,

The only thing that is IMHO not as it "should be" is the star coloring, for me it is to greenish, try  some SCNR to green or perhaps tighten the green channel with curves...
NB images always tend to result in some unnatural star colors..

Best
Title: Re: vdb142 - narrowband - have a try
Post by: astropixel on 2012 May 19 17:45:21
Something like this?
Title: Re: vdb142 - narrowband - have a try
Post by: Lex on 2012 May 20 00:34:33
Absolutely, this one is really nice!
When you create a starmask and invert it to protect the background you will be able to "whiten" up the stars even more!

Nice
Title: Re: vdb142 - narrowband - have a try
Post by: mcbbcn on 2012 November 13 15:14:34
Hi,

As I'm learning about bi-color combination, I took your Ha & OIII and apply the Cannistra method...now, I intend to figure out how to do it in Pixinsight...but this is the result so far using PS...more to come...

By the way, I know it's overstretch, but I was more interested to see the color balance outcome...

Cheers,

M.
Title: Re: vdb142 - narrowband - have a try
Post by: starhopper on 2012 November 14 04:06:03
Hello Herbert,

Quote
On the one side is the mixture of the channels with pixelmath

sorry I must say, I have no solution. To say it in clear words, we all need somebody who provides a mechanism to imitate Photoshops "selective colors".

My workflow:

PI:
HistgramTransformation with STF (auto) to all channels
RGB with Pixelmath: R=SII, G= 0,85xHa, B=OIII

PS: selective Color (to get the Hubble Color Feeling) and dynamic

PI:
StarMask (Scale=8, Smoothness=7, Small=2)
MT (Morphological Selection, Iter=10, Amout=0.15, Selection=0, Sitz 5)
2x Unsharp Mask (Std=0.8, Amount=0,75x)
JPG Export
Title: Re: vdb142 - narrowband - have a try
Post by: Herbert_W on 2012 November 14 04:42:37
Hi Forum, Servus Thomas!

In the meantime, I gave it up, to find a standard routine.
But new Scripts are available - see: http://pixinsight.com/forum/index.php?topic=4853.0 (http://pixinsight.com/forum/index.php?topic=4853.0)

Maybe, these Scripts makes an end to an undless jouney in the sho-universe.
But nevertheless, Photoshops "selective colors" will be the indicator.

> Thomas: thank you for image-processing and the parameters. I never get such intensive and bright colors with PI - there is still a lot to learn!

Thank you and best regards.

Herbert, Austria
Title: Re: vdb142 - narrowband - have a try
Post by: mcbbcn on 2012 November 15 22:13:39
Well, I tried to simulate the Cannistra method in Pixinsight, and I don't think it came out too bad, your thoughts are welcome...this is just a Ha & OIII combination...I was quite surprise, as you can see yellow stars, blue stars, red stars, and there is a bit a pretty good color balance...

If somebody asks me to share the method, I'll be embarrassed  :-[ because as Sanders pointed out, there is more alchemy to the Cannistra method than science...but hey, it's not too bad if you're looking for colors that a lot of people are used to see in OSC shots.

Cheers,

M.

Title: Re: vdb142 - narrowband - have a try
Post by: starhopper on 2012 November 16 09:04:00
Hello M.
Quote
If somebody asks me to share the method, I'll be embarrassed
That's what I do. Can you provide a screen copy of your personel pixelmath formula?

I have successfully made bicolor Images with Cannistras workflow in PS.

Cheers Thomas
Title: Re: vdb142 - narrowband - have a try
Post by: mcbbcn on 2012 November 16 13:55:18
Well, I guess it was meant to happen.  Now, I'm embarrassed...  :-[

...but I'll do more than put the pixelmath values and I'll attach an RTF doc that contains the flow and the pixelmath values.

Just to give you some background on how I went about it.  I took a regular daylight image in PS, and I applied Cannistra's color balance.  Then, at the same time, I opened Pixinsight, and I started to play around with a Pixelmath formula that Juan just shared with me to control shadows, midtones & highlights.  So, when the Pixinsight image look close to the PS image, I annotated the values.

Then Cannistra uses blending PS modes, and as Gerald has a great page with the PS Blending modes formulas for Pixinsight & other programs, it was relatively easy to extrapolate some of the blending modes in Pixinsight using Pixelmath.

As you will see in the instructions, I had to do a bit of tweaking with Curves Transformation, but in the end, it came out pretty good color-mapping wise for an Ha & OIII bicolor combination.

My documentation is not elegant in any shape or form, and the Pixelmath formulas, some...I hard-coded the values, and some others I used variables...I'm sure it can be done in a much better way.

I spent quite a lot of hours doing trial & error to get the as close color balance as I could, but if you find a better way to do it, please post the results.  In the end, color mapping is like food & flavors, everybody is going to be right.  :)

See link to download the document: http://dl.dropbox.com/u/8971416/Cannistra_pix.rtf

Enjoy it!

M.