Siril Initial Post Processing Workflow

I'll describe here the steps I follow to get from the stacked Linear image result from the Pre-Processing workflow to an image which could be distributed, although progressing to generating a Starless version and possibly refining the image with Photoshop would definitely produce a better final image. Obviously for this you will need the Siril application, download for your platform and install. In many cases the process below is following along with a tutorial provided by the Siril team, however I do make some recommendations along the way and will point this out. Please use this article in conjunction with the Siril Tutorial.

Summary

In summary these are the steps I take from the Pre-processed Stacked Linear image (end of Pre-processing): -

Following on from this you can optionally move to separation of the stars from everything else Or skip to Post Processing in Photoshop. Actually, a controversial point of view here is that you could stop processing at this point and it is more than likely it would be good enough to post to social media with the that platforms own prebuilt filters.

Siril - Crop

Cropping is NOT optional! You might have correctly framed your image and all your lights look precisely as you wanted but at the very least there will be artefacts of the registration and stacking process that look bad. In many cases you can improve on the framing of the object for dramatic effect or to emphasise a specific aspect of the target. Of course be a little careful, as those photons that lit up these pixels you are discarding took some capturing!

The screenshot opposite shows the Siril Cropping tool. Simplicity itself, drag a rectangle you wish to keep, right click and choose Crop. You have to do this on the R,G or B channel. More detail as part of a Siril Tutorial.

Siril - Rotate

As part of framing you may have an idea of the orientation of you target, you may have been a bit lazy and centred the target without much thought about orientation (after all there is no up in space!) or it maybe that some objects have just become known to be orientated in a certain way. Whatever the reason Siril is able to easily rotate your image in 90 degree increments or in any precise angle you might be wish. Purely optional but do you really want an upside down Lion nebula!

I turn off the crop option as I want to do this manually if I feel it is needed, which is unlikely as I cropped in the previous step.

Siril - Background Extraction

This is another optional step it may be necessary if you took the decision not to remove gradient at the pre-processing step from each light frame OR there may still be some gradient you wish to remove.

If you proceed wit this then you are generally dealing with a complex gradient as you will have brought in gradients from each of the light frames and therefore there will be rotation of those gradients through the night as the camera tracks the target. We therefore need to pick an appropriate algorithm to use in Siril to allow it to find the best fit. Earlier version of Siril would only provide the Polynomial algorithm and it would be suggested to use a 4th order polynomial. More recent versions and therefore the option I use now is shown, this is an RBF method. I tend to leave the defaults in place, click the Generate button and it will place the red dots across the image, these are the sample points it is going to use for the background. The Grid Tolerance can be adjusted and Generate hit again to make sure that none of the sample points are on the nebulosity or bright area of the target. It usually does a fairly good job at missing but it is sometime necessary to remove some sample points manually. This is done by right clicking, you can also add points by left clicking. Once done hit Compute Background and then Apply (make sure that apply to sequence is NOT selected).

Siril - Photometric Colour Calibration

I follow the steps in the manual processing tutorial on the Siril site for this, but why is this step necessary? Read on!

What we have caught and stacked so far are the colours that our camera has seen - this is wrong! Camera sensors for One Shot Colour cameras and this includes Astro cameras use consumer sensors designed for mobile phones , cctv etc... they are not specifically designed for astronomic use, doing so would make them prohibitively expensive to us amateurs. The side effect of this is that generally these OSC sensors have been designed to work in daylight with reasonable amount of signal, i.e. normal photography during daytime or even a bit of low light. The sensor therefore captures what it thinks is a good rendition of what the human eye would like to see in those conditions. Human eyes in daylight are very sensitive to green, we live on a green planet it makes sense for our eyes to have involved to detect differences in lots of shades of the colour we are surrounded by in nature. Enter the bayer matrix, this is a filter that separates each pixel in an OSC into Red, Green and Blue Sensitive pixels. To approximate the image seen by our eye the camera has 1 red, 1 blue and 2 green pixels arranged in a square. This special Colour Filter Array (CFA) is the de facto standard for colour sensors. My 24mp camera will have 6mp of red, 6 of blue and 12 of green. The outshot of this is the colours our cameras see at very low light levels has a tendency to have green cast, in fact it is very unusual in deep sky objects to find green at all - there are no green stars for example!

To counter this there are two options that involve with setting the true colour, Photometric Colour calibration is the easiest and the most scientifically accurate as it uses accurate colour data captured by professional astronomers across thousands of stars. In Siril, we tell it wwhat are of sky we have taken an image of, it will use a technique called plate solving to precisley identify the stars in the frame and then obtains the real colour data, a calculation is then made to produce transforms to bring the red, green & blue channels of our image into alignment with the real colours. When this is applied the stars of course are more accurately represented but so are all other objects in the image.

Siril - Deconvolution

I follow the steps in the manual processing tutorial on the Siril site for this, but why is this step necessary? Read on!

Stars are a long way away, obvious of course. Given their distance no matter what focal length a star would never be seen as a disk in your pictures, but sure enough if you zoom into one of your stars in your images it will present covering a number of pixels. Atmospheric disturbances and imperfect optics cause this haziness, and of course if it is happening to the stars then it is also happening to the clouds of dust and galaxies we are also capturing. I have read that these imperfections are convolving with the the true image hence the term to remove this is to perform Deconvolution. There is a lovely description of how this works and then a description of how to perform this in PixInsight on astrodoc's site, Siril has what I think is a less sophisticated version. My Before and After version can be seen on a very close up area of an image. Deconvolution must be done on Linear data (as we are doing here).

In essence deconvolution is aiming to bring out detail in the image, in doing so it also has the effect of reducing the size and glare of the stars so you get a less star dominated image with enhanced detail of the wispy clouds and nebulosity. One caution, it is easy to over process, when this happens you will end up with little donuts rings around stars which obviously looks ugly - be careful.

Siril - asinh Stretch

So all the previous steps have been working with un-stretched (linear) data. Even when you have been able to see an image on screen in Siril for instance this is likely because you have AutoStretch or Histogram stretch for the viewer turned on. If that is the case then this stretch is only working on what you see on the display, it has not stretched your source data. We are about to change that. In a simple way we are going to permanently change the data so that it reflects the auto-stretching that goes on in the viewer. A great explanation can be found in this blog post on Sky and Telescope. There are many ways to do this, we could for instance perform a Histogram stretch directly on our linear data, in fact that is what I used to do in other software applications. A more refined approach is to first perform a hyperbolic arch sin (asinh) stretch, iteratively so you can just see the stars and nebulosity coming through. Doing this before Histogram stretching preserves the colour and reduces excessive saturation, the asinh algorithm stretches data that is sympathetic to the way the human perceives variation in colour intensity.

My approach is very similar to the Siril tutorial on this subject although I tend to be less aggressive with Stretching in Siril. In a later step we further manipulate the stretch by using levels and curves in Photoshop which I feel allow ore refinement, so treat this as a first light touch: -

Ensure that you are viewing the image in Linear mode and that the two slides are fully to the right and the left as can be seen in the first screenshot. Zoom in to roughly the middle of the image, you will want to see the effects on stars as we go through he process. The first screenshot was around 300%.
From the Image Processing menu select Asinh Transformation... I prefer iterating over this rather than going for one specific big transform - I find I get finer control. So in the second screen shot I have 30 as the Stretch parameter. As you can see with the preview on the effect is noticeable - you can start to see stars. 30 seems to be a good default for me, it is likely going to be different for you. Apply.
Repeat step 2, in my case I kept the stretch as 30. Stars getting brighter and have a green tinge to them, not too worried about this but do not want to push this much more, the background was still looking fairly dark at this point so left the Black Point at zero. Apply.
Repeat step 2, in my case I started with a stretch of 10 and the stars stared looking very saturated and large ('cartoon like'). Therefore I backed off to 3.0. Also the background was starting to look a bit bleached black. So experimented with Black Point and ended up with 0.03. I was happy at this so Apply.

The final screen shot shows the effect overall by zooming back out to full screen. Hopefully you should be able to see stars and some nebulosity. At this point I would save the image to a different name, it seems a good point given that we have just moved from our Stacked Linear to a stretched image.

Siril - Histogram Stretch

Histogram stretch is the final stretch we will perform in Siril. Although this can be done in Photoshop I prefer doing it here as it is more automated, however if at any point I feel like the automated method is over processing by virtue of saturated colour or clipping of the data (more on this later) then I will hire on the side of caution and pull back shadows or midtowns to get to a more conservative image, there is more time to perform additional stretching in Photoshop later. I pretty much follow the Siril Tutorial notwithstanding the earlier comments about being conservative: -

Ensure you are in Linear mode with the top and bottom slides fully to the right and left (this should be the case if your are following along). I usually do this at full screen mode and from the Image Processing menu select Histogram Transformation... As per the first screen shot you will see the histogram for your image, it will typcically look very similar with a peak towards the left hand edge, i.e. showing you that most of your data are low level values.
The plus and mini buttons allow you to zoom in and out. Zoom in so you get a good view of the histogram. In my case this is with a zoom of 5. This really only matters as in a later step you may wish to drag the Black and grey arrows around (Shadow pointer and Mid-tones pointer) and being zoomed in gives you more accuracy.
Click the button that looks like a cog with a curved line going through it. This will apply the histogram auto-stretch. Your image should significantly change.
It is now possible for us to alter this automatic stretch by moving the sliders (the triangles on the x axis of the histogram). I do this in two situations:-
- If the clip (%) for shadows is greater than 0.1 (or thereabouts) - in which case drag the Back Point (shadows) pointer to the left slightly and watch the clip %
- If I think the image could be stretched a little more or less, conservative changes of the Grey arrow (mid-tones).

The final screenshot shows the full image after Applying the changes to the histogram stretch window.

Siril - Green Noise Reduction

Noise reduction is one of those subjects that it is difficult to be definitive on, I noice some astrophotographers do not need to do this when they have already performed Photometric Colour calibration, yet some do. I fall into the latter group. It definitely takes the green tinge away from the stars in the image. I do exactly as per the Siril Tutorial.

You can see a before and after just using the Siril defaults accessed via the Image Processing menu and Remove Green Noise...

Another trick I have seen for images with a magenta cast (rather than green) is to use the same function but do this on an inverted image. You can invert the image in Siril using <CMD>+I, I believe this will be <CTRL>+I on Windows. Then run the Remove Green Noise before Inverting your resultant image to get back to the positive. I do not seem to have this Magenta noise to contend with so do not do this.