NARROWBAND IMAGE PROCESSING IN PIXINSIGHT & PHOTOSHOP: PART 1
This two-part tutorial will guide you through my basic workflow that I follow to produce narrowband (Hubble Palette) astro-images in PixInsight and Adobe Photoshop. Although specific processing steps can vary from image to image, the beginner to more advanced astrophotographer will be able to achieve solid results by following and adapting this workflow to their own images.
The following software is required to complete Part 1 of the tutorial:
For both parts, I will be using narrowband data of The Eagle Nebula (M16) obtained from the Telescope Live CHI-1 telescope, however the steps that I will be guiding you through will work with any narrowband data.
This tutorial doesn't cover calibration of raw data and assumes that the data you are processing is full calibrated with dark, flat, bias and/or dark flat frames. It is also assumed that you have a basic knowledge of the operation of the functions in the above software.
Workflow Overview: Part 1
An summary of the basic workflow for Part 1 of this tutorial is given below, however we will be going through each step in detail:
Initial Processing - PixInsight
Cropping - cropping of the Ha frame to use as a reference for the OIII and SII frames.
Registration of Ha, OIII and SII frames - use of the StarAlignment process in PixInsight to register the Ha, OIII and SII frames.
Colour (SHO) Processing - PixInsight & Photoshop
Linear gradient removal - use of DynamicBackgroundExtraction to remove any gradients from the Ha, OIII and SII frames.
Linear noise reduction - use of AI-powered NoiseXTerminator to remove noise from Ha, OIII and SII frames.
SHO combination - combination of SII, Ha and OIII frames to produce a colour image.
Colour calibration - use of AutoColor script to neutralise and calibrate the colour channels in the SHO image. Also, the correction of star colours using PixelMath.
Stretching - use of HistogramTransformation to stretch SHO image from a linear state to a non-linear state.
Non-linear noise reduction - further reduction of any noise in the non-linear SHO image.
Non-linear colour balance adjustments - use of Bob Franke's Hubble Palette Methodology to make further adjustments to the colour balance in Photoshop.
HDR adjustments - use of EZ HDR tool to reduce any overexposed areas in the SHO image and produce a HDR image.
Let's explore these steps in more detail:
Initial Processing - PixInsight
Initially, open your separate Ha, OIII and SII frames in PixInsight and apply an STF AutoStretch to each frame using the STF AutoStretch button:
We will be cropping the Ha image and using it as a reference for OIII and SII frames. Open DynamicCrop by going to Process > All Processes > DynamicCrop and draw a box to produce the desired field-of-view in the Ha frame:
Now click the green tick in the DynamicCrop window to crop the Ha frame:
2. Registration of Ha, OIII and SII frames
We now need to register (align) the OIII and SII frames with the Ha image. To do this, open the StarAlignment process under Process > All Processes > StarAlignment and set the reference image to the Ha frame under View and add the OIII and SII frames to the target images box by clicking Add Views. Leave the rest of the settings at the default. Then click the global button at the bottom of the box to execute the process:
This will produce OIII and SII frames that are aligned and cropped to match the Ha frame exactly. Close the original OIII and SII frames, then rename the new OIII and SII frames accordingly. Apply an STF AutoStretch again to all three images:
Colour (SHO) Processing - PixInsight & Photoshop
1. Linear gradient removal
Now that we've cropped and aligned the three channels, we will now move on and produce a colour image mapped to SHO (Hubble Palette). There are many other different combinations in which narrowband images can be mapped to e.g. HOS, HOO etc. However to keep things simple, I'm going to stick to SHO, which provides an effective and attractive result.
The first step is to duplicate the Ha frame, as we are going to use that later on in Part 2 of this tutorial as a luminance layer to add detail to the image. To do this, simply click and drag the image identifier of the Ha frame away from it to produce a clone, then rename the clone to L as shown below. Then save the image.
Minimise the L frame and move it to the side for later on. We're now going to remove any gradients that exist from the Ha, OIII and SII frames using the DynamicBackgroundExtraction process. To open this go to Process > All Processes > DynamicBackgroundExtraction. Now click on the Ha frame and a cross will appear on the image.
When there is a well defined, dark background in the image, such as when imaging a galaxy, the standard methodology is to place sample points onto the background, avoiding stars and any objects such as nebulosity etc. In this case, you can usually leave the settings at the default, with the exception of the tolerance and shadows relaxation. I'd recommend increasing these to 3.0 and 6.0 respectively. After clicking Generate, this should provide good sample points on the image. It may be necessary to move these around after generating them, as some may fall onto the top of stars, which may impede the effectiveness of the gradient removal. The window in the dialog box provides a good indicator of whether the points fall on top of any stars.
In this case however, the image doesn't really have a clearly defined background, as the nebulosity covers most of the frame. Hence the above method will not work effectively. A good way and a trick I've found of placing sample points onto images without a dark background, is to place them around the edge of the image and then delete those in the centre. I've found this gives good results. Also, in addition to increasing the tolerance and shadows relaxation, in this scenario, I'd also recommend increasing the default sample radius by x10 and reduce the samples per row to 7. A bit of experimentation may be required to get the desired result. Also we need to change the correction type to Subtraction and tick discard background model and replace target image, as shown below:
Once we have these settings, click Generate and the sample points will be placed on the image as shown below:
We will now delete all of the sample points with the exception of those around the edge of the image by simply clicking on them and pressing the Delete key. The image should look like so:
Now click the green tick at the bottom the DynamicBackgroundExtraction window to execute the process. The image will now be corrected and any gradients will be removed or significantly reduced. It may be necessary to apply an STF AutoStretch to the image after DynamicBackgroundExtraction, but the image should now look like the below:
Repeat the above steps for the OIII and SII frames and you should be left with three well corrected images:
2. Linear noise reduction
We will now apply noise reduction to our linear frames using the NoiseXTerminator process. It is nice and simple to use, with only two parameters that can be adjusted; denoise and detail. The default settings of 0.90 and 0.15 respectively work perfectly fine with most images, although feel free to experiment. We simply drag the blue triangle in the window onto the Ha frame to apply the denoise process:
We can see the result below of using NoiseXTerminator on our images; the noise has been significantly and effectively reduced without compromising detail:
Repeat this for the OIII and SII frames. We will now combine our frames to produce a colour image.
3. SHO combination
To combine the Ha, OIII and SII frames into a colour image, we will use the ChannelCombination process and map the channels SHO to produce a Hubble Palette image. Go to Process > All Processes > ChannelCombination and map the frames as shown:
Click the global button in the window and a SHO image will be produced. Rename the image to SHO and apply an STF AutoStretch to it. The image will likely show a strong green bias, due to the strong signal of the Ha channel relative to the others:
We could have performed the LinearFit process to the frames before combining them, which may have produced a better colour balance after combining them, however this is purely at the discretion of the imager.
4. Colour calibration
We will correct the colour balance using the AutoColor script. To do this, go to Script > Utilities > AutoColor and it will correct the colour balance of the SHO image as shown below (again, an STF AutoStretch may be required after applying the script):
This isn't the desired final result however, we still have quite a few more adjustments to make. Firstly the image has magenta stars; a common issue when producing narrowband images. We will correct the magenta stars by using the PixelMath process. Open this by going to Process > All Processes > PixelMath.
Ensure that use a single RGB/K expression is unticked and enter the following expressions into the relevant channels:
I found these expressions in this forum post and they seem to work very well. After entering the expressions, apply the process to the SHO image by dragging the blue triangle onto the image. The magenta tint on the stars has now been significantly reduced:
We're now going to stretch the image from a linear state to a non-linear state.
Reset any STF adjustments that have been applied to the SHO image and open the HistogramTransformation process by going to Process > All Processes > HistogramTransformation. Ensure that the SHO image is selected in the HistogramTransformation window and click on the real-time preview icon at the bottom of the window to allow real-time adjustment of the histogram:
With the real-time preview window open, drag the middle grey slider of the histogram to the left and we can see that the image is becoming brighter:
This is essentially moving or stretching more of the lower signal of the histogram into the higher levels, making the image brighter. Apply the stretch by pressing the blue square in the bottom left of the window and reset the HistogramTransformation process using the reset icon in the bottom right of the window:
Repeat this iteratively until the foot of the histogram is positioned at approximately a quarter the way into the full range, as shown below:
Next adjust the black point slider by pulling it to the right, leaving a bit of space between it and the foot of the histogram. We don't want to clip any pixels in the image. There is a box that shows if any pixels are being clipped next to shadows, highlighted in red below:
Apply this new adjustment and iteratively adjust both the black point and middle grey sliders until the desired stretch is obtained, ensuring that no pixels are clipped at the same time. A lot of this is down to personal taste, but my own final stretch is shown below for reference:
The SHO image is now in a non-linear state. Close the real-time preview and HistogramTransformation window and we'll now move on to non-linear noise reduction.
6. Non-linear noise reduction
Similar to when we performed linear noise reduction, we're simply going to apply the NoiseXTerminator process again to our non-linear image using the default settings. We can see the end result below. Although subtle, there is a slight improvement:
Save the image as a 16-bit TIFF file. We're now going to move over to Photoshop to make some further enhancements to the colour balance using the Selective Color tool.
7. Non-linear colour balance adjustments
This methodology was created by Bob Franke and uses the Selective Color tool in Photoshop to optimise the colour balance of Hubble Palette images. Find out more here.
Open the SHO TIFF file in Photoshop and open the histogram on the right-hand side of the window. We can see that the channels are roughly balanced. This is our 'base' SHO image that were are going to work from:
We're going to convert the base SHO image into one that has more of a characteristic gold and turquoise motif.
Open the Selective Color tool by going to Image > Adjustments > Selective Color. Now adjust the Greens, Yellows, and Cyans sliders accordingly as shown below:
Click OK and we can now see that the image has more of a gold and turquoise motif:
Save the image and we'll return back to PixInsight to perform one final process on the SHO image.
8. HDR adjustments
This is just a personal preference, however I like to reduce the contrast between the darker and brighter areas of my images i.e. increase the HDR, as some of the brighter areas can become blown out. To achieve this, I use the EZ HDR script. Open this by going to Script > EZ Processing Suite > EZ HDR. I find the default settings perfectly adequate for most images, but again, feel free to experiment as creativity is the name of the game here.
Ensure that the correct image is selected in the dropdown list and run the script. We can see the results below; although very subtle, the contrast between the darker and brighter areas of the image have been reduced:
Our SHO image is now complete, make sure to save it! In Part 2, we will process the luminance image and combine it with our SHO image to produce the final composite.
Part 2 coming soon...