StarTools

Hello. I am new to the hobby and have my first set of data. I grabbed the pixinsight demo and tried and got so-so results I guess. Just so not user friendly. But then I heard about this program and wanted to try the demo. In both programs I have these two black spots (on on the left and one on the left). I think maybe dust? I did do flats, flat bias, and darks. I also have a streak of light artifact. Maybe some type of reflection at some point? Anyway I was wondering if I could get some guidance on how you remove areas like that in Startools. I used the wipe function but still keep bad gradients as well as the two dark spots. I watched a couple tutorials and used the lasso feature to mask the spots when doing wipe so they would not interfere. But I get less than stellar results. Here is the link to the .fts file if anyone is willing to help me out. thanks so much

https://drive.google.com/file/d/0B0WZUC ... sp=sharing

If you re-save the file with a different name and in .fit 32bit unsigned integer format, Startools resulting image is less noisy, I don't know why.

OTOH, flats calibration should have taken care of dust donuts. I've previously had problem with calibration so I'd stack uncalibrated files as a reference. Any calibration you do should get a better image. Try these steps and see if your image improves step by step or if suddently something goes wrong:
- stack images
- stack images with darks
- stack images with darks and bias
- stack images with darks, bias and flats
- stack images with darks, bias,flats and dark flats

Regards

waterbourn wrote:Hello. I am new to the hobby and have my first set of data. I grabbed the pixinsight demo and tried and got so-so results I guess. Just so not user friendly. But then I heard about this program and wanted to try the demo. In both programs I have these two black spots (on on the left and one on the left). I think maybe dust? I did do flats, flat bias, and darks. I also have a streak of light artifact. Maybe some type of reflection at some point? Anyway I was wondering if I could get some guidance on how you remove areas like that in Startools. I used the wipe function but still keep bad gradients as well as the two dark spots. I watched a couple tutorials and used the lasso feature to mask the spots when doing wipe so they would not interfere. But I get less than stellar results. Here is the link to the .fts file if anyone is willing to help me out. thanks so much

https://drive.google.com/file/d/0B0WZUC ... sp=sharing

Hi,

Welcome to the hobby! I have to say that is is pretty darn good data for someone just starting out. Before addressing your particular questions, one thing that will further really enhance your data is dithering. Right now you have noise that is a bit 'streaky' in nature. By moving your camera a little between shots (preferrably in an outwardly spiralling pettern) you will get rid of that noise and you'll be able to push it much further. Having noise in your image that follows a 'pattern' (e.g. streaks), really hampers any noise reduction routine's ability to discern noise from detail.

Now for your specific questions. Yes, those two spots are most likely dust. Masking them out like so...
...should help WIpe ignore them. Further gradient related trouble can be addressed by making sure that there are no further (smaller) 'dark' anomalies detectable - dark anomalies are any pixels that are darker than the true interstellar background that was recorded; e.g. simply anything that is artificially 'dark'. If an image is noisy, bumping up the Dark Anomaly Filter settings is usually a good idea. I further found that using the Vignetting preset helped with your data as well, since some darkening towards the corners was still visible.

Further processing may sufficiently mask the dust donuts' presence (especially if they're obscuring interstellar background). However, if they're not simply obscuring interstellar background, or you prefer a level of skyglow in your background. Have a look at this thread showing you how to painlessly remove such donuts from your image. Note that this technique only works for things darker than the 'expected' data (e.g. dust donuts). You can also use the Heal module for healing out reflections if/when necessary, but this starts to enter the realm of 'doctoring' your image (do let me know if you want to know how to do this regardless - as opposed to PI we're not all about pushing arbitrary 'philosophies' here ).

This is an example of what you could end up with.
Processing flow as follows;


--- Auto Develop
AutoDev to see what we got. Default parameters (we'll redo our stretch later and we don't care about the right settings right now).
We can see streaking noise, gradient, the two dust donuts and a red bias. Noise grain is 'clumpy' (due to debayering).
--- Bin
We'll trade resolution for noise reduction. This also undoes (as best as it can) the interpolation that debayering caused, making noise grain 1:1.
Parameter [Scale] set to [(scale/noise reduction 50.00%)/(400.00%)/(+2.00 bits)]
--- Wipe
Created mask with dust masked out. Vignetting preset.
Parameter [Dark Anomaly Filter] set to [4 pixels]
(time traveling thanks to Tracking feature - Wipe applies the gradient removal to the linear data, e.g. before we stretched it with AutoDev)
--- Auto Develop
Time to put the dynamic range that Wipe freed up to good use!
Redoing global stretch (time traveling thanks to Tracking feature, AutoDev goes back to linear data), specifying a ROI (click & drag) over M51A/M51B.
Parameter [Ignore Fine Detail <] set to [6.3 pixels] in order to make AutoDev ignore the fine noise.
Parameter [Outside ROI Influence] set to [3 %] in order to make AutoDev ignore anything outside the ROI even more.
--- Deconvolution
Click on the Auto Mask button to have StarTools create a suitable de-ringing mask for us automatically (the current mask is not suitable - we designed it ourselves for the Wipe module).
Parameter [Radius] set to [2.3 pixels]
Again, time travelling thanks to Tracking feature, Decon goes back in time when the image was still linear, applies the deconvolution, then travels forward in time, reapplying Wipe and AutoDev. If you ever switch to PI, please be aware that deconvolution on stretched data is not useful/mathematically correct!
--- Life
Using the Life module's Isolate preset to push back the noise, while 'Isolating' the DSO. Clear, Invert the mask (e.g. select every pixel for processing) before applying.
--- Color
Final color calibration. Again, thanks to time traveling this is something that can be done anytime in StarTools, preferably towards the end so ST can compensate for you stretching the brightness/luminance and still get correct color. In other, less advanced software such as PI or PS, do this after DBE/ABE/GradientXTerminator and before stretching your data - proper color calibration of stretched data is, like deconvolution, only applicable to linear data.
The default values that StarTools came up with are too green (less noisy data will help ST make better guesses).
Rule of thumb is to look for a good even distribution of all star temperatures (from deep red to blue). A typical (there are many exceptions) galaxy has a yellowish core (older stars) and and a bluer outer rim (younger stars) with purple HII knots/areas spread throughout.
Parameter [Cap Green] set to [To Yellow]
Parameter [Dark Saturation] set to [4.50]
Parameter [Bright Saturation] set to [Full]
Parameter [Saturation Amount] set to [300 %]
Parameter [Green Bias Reduce] set to [1.52]
--- Wavelet De-Noise
Final noise reduction. Tracking will have data-mined your data and processing choices by now and will have pinpointed exactly where noise has been brought out.
Parameter [Scale 5] set to [50 %]
Parameter [Brightness Detail Loss] set to [9 %]
Parameter [Grain Size] set to [22.1 pixels]
Parameter [Read Noise Compensation] set to [11.18 %]
Parameter [Smoothness] set to [66 %]

Hope this helps!

I really appreciate the info! Taking the time to really go through the data. Ir really shows your commitment to your program and wins me over for sure.

I will try these steps and see what I can do. I spent hours playing around with the settings and didn't even get half of the image you were able to get. Knowing what is possible is awesome. So couple quick questions before I try.

I also see the really red gradient when I autodev. What causes that red?

When you do that first auto dev, are you "keeping" it or just seeing what you get, not keeping it and going to the wipe function?

Once you AutoDev then the tracking button is what gets you back to the unstretched image. Here you choose "Linear"?

Also when you first load your image I am unsure if I pick option 1 or 2 (not bayered vs bayered).

And then what step is really getting the dark background? That is where I a was struggling the most. I never could even get close to that darker black background. But in this case the background almost seems too dark because I can't make out the other "fuzzies" that are present when there is much more noise.

I am sure I will have plenty of other questions. Again, I really appreciate your time and dedication to helping this newbie out.

One more question. I autodev (and "keep" it), then binned (and keep), created a mask over dust spots (keep), used wipe feature (keep). But when I do this I still have a very red image, the spots are still there and still a lot of noise. I am kinda lost on the "go back in time" feature. Am I missing a step Should I be hitting the track button or the restore button at some point?

waterbourn wrote:I really appreciate the info! Taking the time to really go through the data. Ir really shows your commitment to your program and wins me over for sure.

I will try these steps and see what I can do. I spent hours playing around with the settings and didn't even get half of the image you were able to get. Knowing what is possible is awesome. So couple quick questions before I try.

I also see the really red gradient when I autodev. What causes that red?

Don't be too concerned with the bias - often it is caused by a background level (such as skyglow or light pollution). And because your color balance is still off at this point, the background level will have a distinct color.
It's Wipe's job to remove as much of this level as possible, without removing actual detail or nebulosity.

When you do that first auto dev, are you "keeping" it or just seeing what you get, not keeping it and going to the wipe function?

Yes, just 'keep' it. It makes things easier to see. There is also the 'Temporary AutoDev' feature in Wipe which will perform a stretch similar (but not the same as) AutoDev, so you can easier see whether there are any remnants or other gradients that need your attention.

It's important to note that it is not Wipe's job to make the background into a nice neutral color (the Color module will take care of that). Especially when noisy, a Wiped background might still exhibit coloring. It's only Wipe's job to free up dynamic range that was otherwise occupied by an unwanted bias/gradient.

Once you AutoDev then the tracking button is what gets you back to the unstretched image. Here you choose "Linear"?

No. The nice thing about StarTools' Tracking feature is that all the modules themselves know what version of the data they require and can 'go back in time' to when the data looked like that (linear, stretched, wiped, deconvolved, etc.). You don't have to do anything. For the user the whole notion of linear versus non-linear data (something that trips up a lot of people in other software) is abstracted away. If you yourself really need to get back to the point where the data was in a particular state, then you can use the Restore button for that. It will give you a set of options to what version of the data it should restore the current image to.

The AutoDev module will, the second time around, ask if you want to re-stretch your the data - AutoDev itself will recover what the image looked like when it was linear. There is no need for you to do anything

Read up on Tracking to find out what the big deal is...

Also when you first load your image I am unsure if I pick option 1 or 2 (not bayered vs bayered).

Pick 1 for now, until you're more familiar with the jargon Looking at your data, it was bayered, but already color balanced.

And then what step is really getting the dark background? That is where I a was struggling the most. I never could even get close to that darker black background. But in this case the background almost seems too dark because I can't make out the other "fuzzies" that are present when there is much more noise.

In this case, taming the background was a combination of a nice global stretch (see AutoDev workflow) that is optimised for the object of interest and ignores the noise around it, and using the Isolate preset in the Life module (also, see workflow).

So I followed the workflow you provided. thanks for that. Here is what I get when I follow it. I still have some red and the dust donuts don't go away like yours. You get a much cleaner background than me with the same workflow. But I do think this program is great. This is still better than what I achieved with the Pixinsight demo (I am sure that someone that knows the program can get wonderful results).

However any other little processing tips would be great as I learn StarTools. Thanks!

Nice!
You probably went a bit too aggressive with the AutoDev ROI (I made sure I could barely see the background). The more background you include in the ROI versus the galaxy, the more it will decide to show background in exactly that proportion (it's pretty cool/intuitive the way it works, but it requires a bit of playing around especially with noisy/uneven data like this). You could also use a manual Develop if you find it easier to get better results with that.

To me, it appears you went a bit hard on the read noise compensation (the stars will start to get blury halos around them when you do) and too light on the Color Detail loss (you retain color noise).

The thing with challenging data is that it is 100x harder to correct acquisition problems in post-processing than it is to fix them at acquisition time... Nevertheless, given the data, that's a pretty sweet result.

You'll find that when you sort out the obvious easy bits (flats, focus, dithering) your data will improve in leaps and bounds, that's even before you start taking longer and more exposures!

I know it sounds a bit weird coming from a 'competitor', but try to make the most of your 45-day PI trial as well. ST isn't so much a commercial enterprise (and I'm pretty confident about how ST stacks up in the post-processing department), so have a look at Harry's videos http://www.harrysastroshed.com/pixinsig ... thome.html for PI. If you're more familiar with one or the other, this thread may help as well.

At the end of the day I'm most interested in getting as many people as possible enthused about imaging the night skies, no matter their means or software used.

Cheers,