PixInsight Forum (historical)
PixInsight => Wish List => Topic started by: mmirot on 2010 January 23 09:04:11
-
PS shop has a selective replacement color tool. I find this very useful for shifting the hues of narrow band image palates. It would be nice addition.
Max
-
Hi Max,
Have you tried using the Curves process?
Cheers,
-
Sure but this different. You are remapping a color selectively. Ie all cyans mapped to a more yellow color etc. See my Soul nebula in the gallery. Both palates were shifted.
Max
-
This may be done with curves. See the h channel.
-
Thanks I give this a try.
Max
-
I gave it a try. I does not work well.
It looks like the right math behind it.
I could not remap cyan areas to blue.
See the image attached.
How can I make the center blue like http://www.imagingdeepsky.com/Nebulae/IC1848/IC1848.htm
This tool :'(
Max
-
Hi Max:
You can separate the RGB channels and then rejoin with LRGB combination, giving more weight to blue.
(http://farm3.static.flickr.com/2725/4373093783_541f66bf58_b.jpg)
After a bit of curves to color and saturation and ¡voilà!.
(http://farm5.static.flickr.com/4013/4373094593_e88da9674a_b.jpg)
Greetings. Silvercup
-
Not sure why you'd split the channels though. You can already work on each channel individually in curves and histograms.
-
Hi Sander:
Try to do curves or histogram and get similar blue tone without lossing greenish and yelowish zones, it's very difficult. So as we want to remap cyan tones to more blue a RGB remap with more blue weigth is an easy solution, but I can be wrong, of course.
Gretings, Silvercup.
-
Thank you both.
I was able to do it in curves. Change the h and Green components. ( H is yellow in curves.)
I guess we can do this just fine with a new tool for shifting hues.
Max
-
This is better
-
Hi Silvercup,
my point is that if you first split the channels, then process them separately with curves and histogram only to combine them again then you're adding two steps that are not needed. Each channel can be processed individually without first splitting it. As an added advantage you can see the effect on the final image while you're tweaking a single channel. Can't do that if you first split it. Naturally you can configure PI to only show the channel of interest, if you prefer.
Perhaps I didn't understand what you were trying to do, sorry.
-
Hi Sander:
I only split an recombine with different weights, I don't do curves or histogram in the splited RGB channels.
I'm writing a new script for Narrowband combination, because when I processed "The Rossetta" it was very difficult to me assigning channel weights to avoid greenish tone due to high HAlpha signal, and reach color tones I want.
So, if you reasign channel weights you can turn a color tone to another one, while the image is still linear, without the need of using curves or histogram.
Of course, you can use the same method with non linear data, althought it isn't the main intention.
I'am not sure if you use H channel in curves you can modify only blue tones, or if green and red tones are affected too.
I'll try to publish script in a few days. I'm applying to my rossetta and I'am very happy with the result, It's much better than my first attemp with trial and error combinations.
Best, Silvercup
-
A script would nice. I was able to achieve a good effect but using the h curve is actual q
-
Silvercup,
A script would nice. I was able to achieve a good effect but using the h curve. Using h is really very difficult.
Hue remapping would benefit from better tools.
Also,
A upgraded channel mixer would be nice for narrow band images. Right now it is tricky formulas in Pixel math.
I would love to see Channel 1, 2,3 etc to be assigned perhaps to a hue. Such as ha channel 1 mapped to Hue 30 degrees ,HSL ( orange?) etc.
The abilty to split the channel would be helpful.
For example, You want 50% of OIII weighted to green and 50% blue.
If some wants to go all the way with this idea it would do four or more channels.
Ha + RGB blends and multichannel narrow band + broad blends would be possible.
I am thinking it could even tackle tasks like Vincent did recently.
Max
-
Hi all,
I have been thinking about this whole subject of 'channel mixing' as well.
In my mind, I visualise a 'grand channel mixer'. This would allow a user to select from a bunch of open images (no real need to work from files I don't think). There would be four 'channels' that will be combined by the LRGB Process, as it stands. No need to re-invent THAT particular wheel. The LRGB Combine process is a 'superset' of the RGB Combine process anyway, so if a user does not want to include anything in the Lu channel, that is fine and simple anyway.
For each of the four 'input channels', the user would then pick appropriate images for inclusion in the blend. And, for each image allocated to a channel, the 'percentage' of inclusion could be adjusted by a slider.
So, by way of example, consider seven source images that a user wishes to 'blend'. Let us assume that they have somehow been obtained, and are now available as:-
Lu
Rd
Gn
Bu
Ha
S3
O2
The user wishes to do a four-channel blend, and decides to try the following
Lu = 50% Lu + 50% Ha, with a final contribution of 30% to the LRGB blend
Rd = 50% Rd + 50% Ha
Gn = 30% Gn + 70% S3
Bu = 30% Bu + 30% Ha + 40% O2
The appropriate images would be allocated to each 'master channel', and the percentage sliders adjusted accordingly
The final combine would reduce the overall Lu combination to 30%, and the LRGB process would be invoked - producing an output image for the user to assess.
However, further options, at the 'channel blend' stage, might allow the user to maintain the 'shape' of the channel histogram, by referencing the channel to a selected image's histogram. That way, the individual Minimums, Medians, Maximums etc, for each 'blend image' in a channel could be adapted such that the overall histogram SHAPE remains 'similar' to the 'shape' of the referenced image - a form of 'normalising' if you like - applied on a channel by channel basis. I see that this would help to keep colour 'balance' under control, even though colour 'content' could be changing significantly (but I may be wrong!!)
I used this approach recently when I multiplied my pre-processed M1 Crab Ha channel with my pre-processed M1 Crab Rd channel (which had been extracted from an image that I had 'worked on' in RGB space, starting from a standard Rd+Gn+Bu dataset). This meant that I was adding back in a modified Rd component that wasn't, overall, any 'redder' than the original extracted channel. It had 'more detail' (due to the included Ha data), and the detail was available in a different 'spatial position' on the image, but the Median position and general 'width' of the Histo curve were, more or less, 'in the same position' as for the original Rd only channel.
Now, I may be completely wrong in this approach - I haven't had enough raw images to work with - but certainly, to me, it 'seems' to be a valid process path to follow.
Does anybody out there have better experience? Carlos? You guys must have to consider this problem in exquisite detail when you are working with your 'professional' data.
Cheers,
-
Right on Niall 8)
Max
-
Hi Max, Nial, Sander:
That is the idea. For the moment you can choose in the script HAlpha Image, OIII Image and SII Image (views). You can assign any RGB color to these views or mark Hubble palette or CFHT palettte for automatic color assignation. You can blend each channel with the other ones, Vgr HAlpha + 25% OIII + 5% SII.
The script do the pixelmath operations with blends channels an then you can assign a independent weight for each channel (HAlpha, OIII, SII).
There are two sliders for STF function. When you hit "Generate" button the script do the operations, STF and try to correct background with ImageCombined-med(ImageCombined), so you can see the aproximate image result.
If you see interesting to add other channels (R G B NII, Etc) I can do.
A hint for a better STF control would be appreciate because Sliders aren't very precise.
Edit:
Because script has intensive pixelmath operations I am not considering Realtime Preview. Script generate a new Imagewindow with the result. ¿Can I open the new window at 0,0?
Best, Silvercup.
-
I was thinking on a different approach. Take each image as if it is a "vector". We choose a particular direction for it (for example, it has to be 550nm, with a given strength or scale). Then, is just a matter of translate that into RGB coordinates, and finally, just add each vector.
Sounds elegant, it may work... but, I'm not sure :D
-
Hi, I just post a preview of Narroband Combination Script on Software Development.
Best, Silvercup
-
Sounds great Silvercup. I'm always interested in finding different ways of doing things.
-
Check out software development section for Silvercup's preview
Max