13 March 2013

TUTORIAL (POST-PROCESSING TECHNIQUE): PixInsight Guide for Narrowband Bicolour Palette Combination

Tutorial Type: Post-Processing Technique - PixInsight Guide

This tutorial is the first in a series of short guide tutorials (as opposed to workflow tutorials). It will essentially only guide you through the actual process of combining Hydrogen-Alpha (H-α) and Oxygen-III (O-III) data in PixInsight for producing a Bicolour Palette image. The tutorial starts with the monochrome H-α and O-III images that have been calibrated with dark, bias and flat images and have been registered with each other so they line up perfectly. My workflow tutorials should be consulted for extensive instructions of post-processing beyond the scope of this guide tutorial.

This tutorial is useful to you if you:

1. Have two narrowband images (commonly H-α and O-III) for an object and want to create a Bicolour Palette image out of them.
2. Are interested in the various Bicolour Palette colour combinations one can achieve.
3. Are using PixInsight for post-processing.

We now begin with the post-processing in PixInsight.



STEP 1. Sorting out our monochrome images

We start with our fully registered and calibrated images (they line up perfectly with each other and have had dark, bias and flat images subtracted from them). 


Here we see both images we will use - H-α and O-III. 

STEP 2, OPTION 1. Colour-combining our monochrome images for Synthetic Green Bicolour Palette

There are various ways for colour-combining narrowband images and the same is true for those seeking to use a Bicolour Palette. A first way to colour-combine these images is to create a true synthetic Green image. This method is extensively documented for application in Photoshop by Steve Cannistra but we will apply this same combination method in PixInsight instead. 

This method involves inserting the H-α image in the Red channel, the synthetic Green image in the Green channel and the O-III image in the Blue channel. The synthetic Green image is essentially created in PixInsight using a special combination of the H-α and O-III images. The special combination of blending modes used in Photoshop by Steve Cannistra can be replicated in PixInsight using its PixelMath tool. Start by opening both of your images (the H-α and O-III images) in PixInsight and opening the PixelMath tool. 




To create the synthetic Green image, insert the following expression into the RGB/K field:
iif(Blend > 0.5, 1-(1-Target)*(1-(Blend-0.5)), Target*(Blend+0.5))
It is important to note that we are essentially blending the O-III image with the H-α image so this means that you should replace Blend with your O-III image filename and replace Target with your H-α image filename. Since my O-III image is called OIII and my H-α image is called HA, I enter the following:
iif(OIII > 0.5, 1-(1-HA)*(1-(OIII -0.5)), HA*(OIII +0.5))
We must of course expand the Destination tab and select Create new image from here. Clicking the square Apply button will then create our synthetic Green image. 



We can now save this image. I saved mine with a filename of sG. We can quickly peak at the difference between this synthetic Green image and the H-α and O-III images by using the ScreenTransferFunction on Auto Stretch. Apply this to all three images. 




Background gradients not withstanding, there are some differences visible between the synthetic Green image and the other two. Mind you, since the H-α image has much stronger signal than the O-III image, the differences are not extremely pronounced at this stage. 

To colour-combine these, we again open the PixelMath tool and click the Reset button to reset the tool. We now unselect Use a single RGB/K expression so we can enter image filenames into the various colour channels. We also expand Destination, check Create new image and select RGB Color from Color space. To use the synthetic Green Bicolour Palette, we will now enter the H-α image filename into R/K, the synthetic Green image filename into G and the O-III image filename into B. Once ready, click the square Apply button. 



You can now close PixelMath, having created our colour image. We can take a peak at the colour balance achieved using this Bicolour Palette by opening ScreenTransferFunction again but this time first clicking Link RGB Channels to toggle it off and then click Auto Stretch



We note that this colour image will need cropping using DynamicCrop, background gradient extraction using DynamicBackgroundExtraction and calibration of colours using BackgroundNeutralization and ColorCalibration but use of these tools are part of my more extensive workflow tutorials and beyond the scope of this guide tutorial. Performing these will show you more clearly the colour balance achieved:


The above result is not perfectly post-processed but it does show you how much difference a background gradient extraction and colour calibration can make to seeing the actual colour balance you have achieved. 

STEP 2, OPTION 2. Colour-combining our monochrome images for LRGB-Style Bicolour Palette

A second way of colour-combining narrowband images for the Bicolour Palette will be what I call LRGB-style. This method essentially tries to mimic what the object actually looks like under LRGB imaging (visible spectrum). We first note that H-α lies deep in the red part of the spectrum whereas O-III lies more on the green part of the spectrum. We can therefore create a Bicolour Palette image using this information by essentially placing the H-α image in the Red channel and the O-III image in both the Green and Blue channels. 

With both the H-α and O-III images open, open the PixelMath tool. Here, ensure you click Reset to reset the tool then uncheck Use a single RGB/K expression, expand the Destination tab, check Create new image and select RGB Color from Color space



As aforementioned, we now enter the H-α image filename in R/K and the O-III image filename in both G and B. Click the square Apply button when done. 



Again we can check our achieved colour balance by opening the ScreenTransferFunction tool, clicking Link RGB Channels to toggle it off and then clicking Auto Stretch




That actually looks more or less like what we expect from the Rosette Nebula as it is very active in H-α and indeed this is therefore a realistic LRGB-style Bicolour Palette. 

We again note that this colour image will need cropping using DynamicCrop, background gradient extraction using DynamicBackgroundExtraction and calibration of colours using BackgroundNeutralization and ColorCalibration but use of these tools are part of my more extensive workflow tutorials and beyond the scope of this guide tutorial. Performing these will show you more clearly the colour balance achieved:


Again, the above result is not perfectly post-processed but it does show you how much difference a background gradient extraction and colour calibration can make to seeing the actual colour balance you have achieved. 

STEP 2, OPTION 3. Colour-combining our monochrome images for Blended Green Bicolour Palette

A third way of colour-combining narrowband images for the Bicolour Palette is vaguely similar to the first presented in this guide tutorial. We again note that H-α lies deep in the red part of the spectrum and O-III lies more in the green part of the spectrum. We will say O-III then lies more toward the blue part of the spectrum (than H-α, which it does) and create a blended Green. This is not the same as the synthetic Green created earlier, but more of a simple blend consisting of a percentage mix of H-α and O-III to comprise the Green channel of our colour image. As we will see, it achieves a different colour balance that is worth exploring. 

With both the H-α and O-III images open, open the PixelMath tool. Here, ensure you click Reset to reset the tool then uncheck Use a single RGB/K expression, expand the Destination tab, check Create new image and select RGB Color from Color space.



We now enter the H-α image filename in R/K and the O-III image filename in B. In G however, we enter a percentage mix of the two. We aim to keep the sum of the percentage mix to 100% to avoid saturating data and clipping it to white. For example, to mix 40% H-α with 60% O-III, we enter the following into G:
(0.4*HA)+(0.6*OIII)
 The following shows this combination:



Again we can check our achieved colour balance by opening the ScreenTransferFunction tool, clicking Link RGB Channels to toggle it off and then clicking Auto Stretch.




We can of course try other blends to get a different colour balance. For example, for 80% H-α with 20% O-III, we enter in G:
(0.8*HA)+(0.2*OIII)
This results in the following:


Conversely, for 20% H-α with 80% O-III, we enter in G:
(0.2*HA)+(0.8*OIII)
This results in the following:


We of course again note that this colour image will need cropping using DynamicCrop, background gradient extraction using DynamicBackgroundExtraction and calibration of colours using BackgroundNeutralization and ColorCalibration but use of these tools are part of my more extensive workflow tutorials and beyond the scope of this guide tutorial. Performing these will show you more clearly the colour balance achieved (top to bottom, 50% H-α with 50% O-III, 80% H-α with 20% O-III and 20% H-α with 80% O-III):




Overall results can indeed be similar but this is expected since all we have really balanced is the Green channel, nothing more.

STEP 2, OPTION 4. Colour-combining our monochrome images for Blended Channels Bicolour Palette

We can certainly extend the above blend of narrowband data to produce a fourth way of producing results with the Bicolour Palette. Here, we still keep in mind that H-α lies deep in the red part of the spectrum and that O-III lies in the green part of the spectrum (tending toward the blue part, at least in relation to H-α). We will therefore blend H-α and O-III in both the Red and Green channels (in different proportions) but keep the Blue channel occupied purely by O-III.

With both the H-α and O-III images open, open the PixelMath tool. Here, ensure you click Reset to reset the tool then uncheck Use a single RGB/K expression, expand the Destination tab, check Create new image and select RGB Color from Color space.




We now enter a blend of H-α and O-III in R/K. Since H-α is deep in the red part of the spectrum, this should form the largest component of the percentage mix without going over a total of 100% (to avoid saturating data to pure white). We therefore enter the following for 90% H-α with 10% O-III:
(0.9*HA)+(0.1*OIII)
In G however, we produce another blend of H-α and O-III. As O-III is more toward the green part of the spectrum than H-α, O-III should form the largest component of the percentage mix without going over a total of 100% once again. We can enter the following 10% H-α with 90% O-III:
(0.1*HA)+(0.9*OIII)
In B we then just enter the O-III image filename. The following shows this combination:



Again we can check our achieved colour balance by opening the ScreenTransferFunction tool, clicking Link RGB Channels to toggle it off and then clicking Auto Stretch.




We again note that this colour image will need cropping using DynamicCrop, background gradient extraction using DynamicBackgroundExtraction and calibration of colours using BackgroundNeutralization and ColorCalibration but use of these tools are part of my more extensive workflow tutorials and beyond the scope of this guide tutorial. Performing these will show you more clearly the colour balance achieved:


This blend looks somewhat different to the previous blends where we only mixed H-α and O-III within the Green channel (it looks somewhat redder). Feel free to play around with the percentage values of H-α and O-III for both the Red and Green channels to produce your desired result.

Kayron Mercieca

No comments:

Post a Comment