Siril osc stacking script

[bomberbaz]

I recently downloaded the one shot colour stacking script, it is a massive time saver for me and tend to use it. I am preferring it as my go-to now but have a query.

If I grab multiple exposures data at eg 120 seconds with ha filter and 30 seconds at luminance, will it pick up both sets of data and stack or just one?

Tia

Steve

[powerlord]

Dunno about script, but if you download the sirilic gui you can specify the seperation easily.

[ONIKKINEN]

For different datasets you need another approach. You can stack them to one image if you want to, but not by using the stock script, or at least not in a way that properly calibrates everything.

What you need to do is to calibrate all the different datasets with their own matching sets of calibration data and then stack it all manually afterwards.

You can use the stock script for calibration, since it blurts out a bunch of files with the prefix pp_light. These are the calibrated files you want to save to another directory for later use (you can delete the rest to save space for now). Do this for any number of nights worth of data and then manually import all of these into a new sequence, then register them and stack in the end.

You can also modify the stock script by simply removing the lines towards the end of the script that register and stack the images. This way you dont have to sit through the registration and stacking phases if all you want is to calibrate.

Or like above, use sirilic which will do anything you want it to but without the manual directory and file moving work.

But you probably shouldnt stack Ha and Lum to one image if that was what you were planning on, this was just a tip on multiple night/different exposure type work.

[bomberbaz]

On 18/10/2023 at 17:41, ONIKKINEN said:

But you probably shouldnt stack Ha and Lum to one image if that was what you were planning on, this was just a tip on multiple night/different exposure type work.

Yes the idea of modifying the script had passed my mind but given the above, where do I go from there? 

What program would you use for stacking the data as one or is there another process you could suggest. 

I mean I could use a universal exposure but this leads to other problems too!

This (stacking data) is a science in itself and I am struggling with some of it,  well actually everything 😭🤣🤣

[bosun21]

You are not alone. I am just about to start some planetary imaging/EAA with the option of perhaps moving onto a bit of AP. I was reading up and watching numerous tutorials on the various software I will be required to master. After a while I was overloaded and slightly overwhelmed by the amount of information I would need to ingest. I am just going to take baby steps and hopefully progress with it all. If it all implodes then I can always go back to 100% visual.

[ONIKKINEN]

50 minutes ago, bomberbaz said:

Yes the idea of modifying the script had passed my mind but given the above, where do I go from there? 

What program would you use for stacking the data as one or is there another process you could suggest. 

I mean I could use a universal exposure but this leads to other problems too!

This (stacking data) is a science in itself and I am struggling with some of it,  well actually everything 😭🤣🤣

For calibration, you can still use the stock script, or modify it a little bit. You can find the scripts in the scripts folder of your Siril installation folder, for my installation it is in C:\Program Files\SiriL and is probably for you too. The .SSF files can be opened and modified with word pad. The simplest way to modify is to just remove the lines at the end, like below:

Just remove those lines and the script ends with debayered and calibrated files with the prefix pp_light for each file. Move those files to another folder and then just delete all of the other files in the working directory. They are temporary files so nothing lost (just make sure you dont delete your actual raw data, keep it elsewhere) and then re-do the script with another dataset with its own calibration frames.

After calibrating any number of images from any number of nights, you will need to now stack manually. Its not too difficult to do, i think its just poorly explained (if at all) by Siril and is pretty confusing for someone using Siril for the first (or tenth) time. What you need to do is import all the files you want to stack into the "Conversion" tab. You dont have to worry about calibration anymore since you already calibrated all the data with the script before. Then in the "Registration" tab you will register (star align) the images, you can use the "Global Star Alignment (deep-sky)" method. Leave the algorithm as Lanczos-4 and interpolation clamping ticked. After registration completes you can now stack the data in the stacking tab. I would recommend the below settings for stacking:

Now for mixing Ha to an RGB image (assuming you are planning on shooting more than Lum and Ha that is) is not so simple. In Siril, there is no couple-click tool solution for that at the moment, but there are some ways to do that. Mixing Ha to RGB is not too difficult in a general purpose processing suite, like Photoshop but not sure you have it?

If you wanted to only use Siril, you would first stack all of your individual filter datasets to their own stacks and then combine the red and Ha stacks with the pixel math tool. If you are working on OSC data, extract the red from the RGB stack. You would first need to register the 2 together, and then linear fit the Ha stack to the R stack to equalize their levels. Then you can use the "max (x,y)" function to combine the 2 images so that only the pixel maximum values from both is left, which will mean that the regions of the image where Ha is stronger than broadband Red will get boosted. This will also boost your noise because the Ha stack will be much noisier in the background, so this is far from a one click thing. You can then blend that to the original Red channel by for example stacking with average and no rejection. If the pixel math wizardry sounds like a bite too much you can also try a simpler 50/50 blend by just stacking the red stack to the Ha stack, which when average stacking with no rejection is chosen will just take the average of the 2 images. This will hurt broadband, and boost Ha. Probably the simplest method in Siril.

Really not sure i would recommend mixing Ha to RGB in Siril for a beginner. Actually having done that a few times myself i still dont want to use that, because there are simpler tools (like Photoshop and PixInsight, but both costly of course).

I'd be happy to walk you through the manual stacking process if there are some head-scratching moments when/if you do decide to try it. The initial hurdle of abandoning the automatic script stacking is the biggest step, but once you get used to doing that it becomes not too cumbersome to do.

[bomberbaz]

20 hours ago, ONIKKINEN said:

For calibration, you can still use the stock script, or modify it a little bit. You can find the scripts in the scripts folder of your Siril installation folder, for my installation it is in C:\Program Files\SiriL and is probably for you too. The .SSF files can be opened and modified with word pad. The simplest way to modify is to just remove the lines at the end, like below:

Just remove those lines and the script ends with debayered and calibrated files with the prefix pp_light for each file. Move those files to another folder and then just delete all of the other files in the working directory. They are temporary files so nothing lost (just make sure you dont delete your actual raw data, keep it elsewhere) and then re-do the script with another dataset with its own calibration frames.

After calibrating any number of images from any number of nights, you will need to now stack manually. Its not too difficult to do, i think its just poorly explained (if at all) by Siril and is pretty confusing for someone using Siril for the first (or tenth) time. What you need to do is import all the files you want to stack into the "Conversion" tab. You dont have to worry about calibration anymore since you already calibrated all the data with the script before. Then in the "Registration" tab you will register (star align) the images, you can use the "Global Star Alignment (deep-sky)" method. Leave the algorithm as Lanczos-4 and interpolation clamping ticked. After registration completes you can now stack the data in the stacking tab. I would recommend the below settings for stacking:

Up to this point I am fine and it is easy enough to follow, I have done script editing so this is all stuff I am familiar with. 

It is the stuff below where we are getting into pixel math where my head starts to hurt.

However, I think I shall give it a whirl first before I start asking for further help as I am keen to learn new methods.

Above you mention that this is a relatively simple operation to carry out with [photoshop, do you know if similar functionality is available within gimp or not?

If you have a video link to the photoshop version I may be able to translate it over to Gimp!

Cheers and thanks for the help thus far, very much appreciate it.

Steve

[ONIKKINEN]

On 23/10/2023 at 22:17, bomberbaz said:

Above you mention that this is a relatively simple operation to carry out with [photoshop, do you know if similar functionality is available within gimp or not?

If you have a video link to the photoshop version I may be able to translate it over to Gimp!

Im afraid i have no video to link as ive not seen one for the subject. Im using a workflow that @ollypenrice shared here on the lounge in a thread not too long ago. Cant find that thread now though, but the basic gist of it is:

Process both the RGB image and whatever image you're using for an Ha image (like a duoband, L-extreme, or whatever you are using) to a point where they have a similar level of stretch and open both as layers in Photoshop. Put the narrowband image in the blend mode "lighten" and create a layer mask for that. Create the layer mask in a way that only the interesting parts get put through, for example by selecting with the color range tool or by simply copying the image to the layer mask.

Just tried this with GIMP and a similar workflow does work. You can put the narrowband image on top of the RGB image as its own layer and put that to the mode "lighten only", which seems to be the exact same as Lighten in Photoshop. From the Select menu choose by color and click on the background to choose that color. Then invert that selection and you are left with a selection that includes everything but the background - so all the interesting parts you want to have. Now you can easily adjust the strength of the process by either modifying the layer mask itself (by clicking on it with alt+left click) or by modifying the narrowband image itself, by curves, contrast, levels, so all the usual stuff. Seems to be fairly straight forward. Definitely more so than the pixel math route which often does not do what you want it to.

For an even better result you could separate the stars from the image with Starnet, and apply the narrowband only to the starless versions of both images and then recombine later with just the RGB stars.

Edited October 24, 2023 by ONIKKINEN

[bomberbaz]

@ONIKKINEN many thanks for this, I shall have a bash at this with what data I have and see what I come up with.

In the interim I found some excellent tutorials on YouTube by one guy who covers a broad range of subjects and goes through them with a Janet and John style for relative novice users to follow. My only problem is trying to remember them all🤣

[ollypenrice]

16 hours ago, ONIKKINEN said:

Im afraid i have no video to link as ive not seen one for the subject. Im using a workflow that @ollypenrice shared here on the lounge in a thread not too long ago. Cant find that thread now though, but the basic gist of it is:

Process both the RGB image and whatever image you're using for an Ha image (like a duoband, L-extreme, or whatever you are using) to a point where they have a similar level of stretch and open both as layers in Photoshop. Put the narrowband image in the blend mode "lighten" and create a layer mask for that. Create the layer mask in a way that only the interesting parts get put through, for example by selecting with the color range tool or by simply copying the image to the layer mask.

Just tried this with GIMP and a similar workflow does work. You can put the narrowband image on top of the RGB image as its own layer and put that to the mode "lighten only", which seems to be the exact same as Lighten in Photoshop. From the Select menu choose by color and click on the background to choose that color. Then invert that selection and you are left with a selection that includes everything but the background - so all the interesting parts you want to have. Now you can easily adjust the strength of the process by either modifying the layer mask itself (by clicking on it with alt+left click) or by modifying the narrowband image itself, by curves, contrast, levels, so all the usual stuff. Seems to be fairly straight forward. Definitely more so than the pixel math route which often does not do what you want it to.

For an even better result you could separate the stars from the image with Starnet, and apply the narrowband only to the starless versions of both images and then recombine later with just the RGB stars.

That's actually more complicated than my method. Rather than layer mask my NB when adding it to the chosen colour channel in blend mode lighten, I give it the kind of stretch which I think will make it play well over the colour channel. I go for a very, very hard, contrasty stretch and ignore noise where it is darker than the same region in the colour channel. It won't be applied anyway. I make sure my background is not brighter than the colour background and that the bright features in NB are brighter than the colour (or they won't do anything.) I adjust this stretch of the NB while it is in situ over the colour in BM Lighten so I can blink it on and off to see what it's doing.

In short I see a NB image to combine with a colour channel as a completely different thing from a nice standalone NB image for publication.

Your NB masking seems a sound idea, though.

Olly

Edited October 25, 2023 by ollypenrice
Clarification