RGB Combination with Pixel Math

[dbennett1]

Greetings -

When combining images from grayscale to RGB with Pixel Math, should those images be in the linear state or non-linear.  Any insight would be appreciated.  Thank you.

Dave

[bulrichl]

Hi Dave,

I am not sure to understand your question right. The combination of grayscale images to a RGB image is done with the process ColorSpaces/ChannelCombination, not with PixelMath.

The grayscale images should be linear, the combining be performed after BackgroundModelization (ABE or DBE) and (if applicable) Deconvolution, as far as I know.

Bernd

[msmythers]

Bernd

Some people use PixelMath for combining color channels.

Dave

As far as to when that is up to your processing flow and what your are trying to accomplish. As Bernd has pointed out most of the time you would want a combined image for things like DBE and whatever method of Color Calibration your going to do.   


Mike

[David Bennett]

Thank you all for the replies.  I was trying to keep the question short and simple but maybe i need to clarify.  I have 3 narrow band channels (Ha, SII and OIII) that have been calibrated, aligned, integrated and cleaned up with ABE,etc.  I was using the linear fit routine with RGB channel combination (all with linear images) but decided to use pixel math so I could combine the 3 channels with some additional control.  For example:
R=0.2*Ha + 0.8*SII
G=0.7*Ha + 0.3*OIII
B=OIII
Prior to combining the images in Pixel Math i moved the images from the linear state to non-linear with Histogram Transform and applied some slight noise reduction. 
I'm wondering now if i should have left the images in a linear state and then converted to non-linear after Pixel Math.  There is a fair amount of discussion on using Pixel Math to combine images, but it's not clear (at least not to me) if the images should be linear or not to get the best color balance.
Thanks again...

Dave

[bulrichl]

Thank you, Mike and Dave for the explanation. Yes, as I never did narrow band imaging I did not think of this option and was completely fixated on RGB broad band imaging. Excuse my ignorance, gentlemen. 

Bernd

[msmythers]

Now problem Bernd. One of the great things about PI is there are more then one way to do things for almost every process. If one was a PixelMath master like Juan and Gerald and so many others in the forum you could do a complete processing with PixelMath. Some people actually enjoy that. Luckily for some of us we have nice programmed GUI's to do the heavy lifting.


Mike

[msmythers]

Dave

There is no hard and fast rule on when to combine. Since I only process an image or 2 of narrow band a year I can't help but maybe rob or RickS or some of the others can chime in. I know for my own RGB work I do things different from the norm like color saturation, star brightness reduction and HDR and all noise reduction work before stretching. That just goes against all normal logic but with PI it is possible. It's what works for me and I'm used to it. I also spends hours experimenting with PI tools just to see how far they can be pushed.

The are also some good postings from a number of years ago on what processes work best on linear/non-linear and either. Hear's one from 2012. https://pixinsight.com/forum/index.php?topic=4055.msg28531#msg28531   That might help with your workflow.


Mike

[RickS]

For true RGB where you want accurate/real colour you should combine and do colour calibration while linear.  For NB you can do whatever you like, IMO!  You should be able to get a pleasing result either way.

Cheers,
Rick.

[David Bennett]

Rick, Mike and Bernd - Thanks a lot for the replies. 

Dave