Autostakkert Quality Graph meaning resolved

[Startrek]

After years, I finally came across an article in Sky and Telescope magazine ( Sept 2016 ) which explains the meaning behind the Quality Graph in Autostakkert 2 and 3
Written by the AS developer Emil Kraaikamp.
The graph shows the following -
Gray jagged sawtooth graph lines represent the video frame by frame unsorted from left to right
Green line represents the frames sorted in quality from best to worse ( left to right )
Vertical Y axis is divided into 4 sections 25% each ( total 100% ) with the 50% quality mark indicated on the graph in the centre
So the best frames are top left on the graph and worst are at the bottom right of the graph
Once the frames are Analysed, you can select the best 50% by holding down the CTRL button an right click with your mouse on the 50% line or roughly where the green quality line passes across the 50% line. This is a safe estimate of quality but Autostakkert recommend you review your frames and based on that try other percentages higher or lower

In the past I have just been manually selecting 25% , 50% , 75% and 90% in the check boxes without fully understanding the graph properly
Now I just select the 50% mark on the graph or roughly where the green quality line passes the 50% line and my stacked images are working out pretty good so far and not waisting time looking at 3 or 4 other stacked options using the check boxes

Hope this helps other folk like me who did not totally understand the quality graph

[Merlin66]

Martin,
Once it's understood it makes things easier.....
Another point: To determine the quality rating of the individual frames, the "Global setting" just picks the "best" whole frame image and sorts them..
Using the Alignment points (ie when imaging the solar disk) it looks at each individual "cell" within the frame ( the 24/ 48 pixel etc.) and sorts the cells into best to worse then recreates the final image based on these cells.

[bojan]

Link to article is here (https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&cad=rja&uact=8&ved=2ahUKEwid0f3Jj-_rAhW6zDgGHVA-AE0QFjAKegQIARAB&url=https%3A%2F%2Fwww.astrokraai.nl %2Fsoftware%2FSky%2520%26%2520Teles cope%2520-%2520September%25202016%2520%5B68%2 520-%252072%5D.pdf&usg=AOvVaw1X9GSfrJ_HQeZmezNX2vYu).

[Startrek]

Thanks Ken and Bojan
I should have shown the link Drrr.....

[Stonius]

Wow! I usually go with far fewer, like best 5%.


Markus

[Dennis]

Thanks for the heads up Martin and the link Bojan.:thumbsup::)

Cheers

Dennis

[Merlin66]

More frames in the stack, the better the SNR.
It really depends on the possible movements (within the target) across the total amount of frames.
Consider a solar prominence.....how much internal movement across 2000 frames vs say 500.
The quality measure is based on the quality of the total frame count- you could have good quality at frame 100 and frame 900 but with different image content.
That's the compromise. I now do my solar in 500 frame videos to minimise the effect.

[Startrek]

I’ve been imaging the planets with my humble Canon 600D and newts over 3 years, back when I started I use to capture 2000 to 2200 frame videos , now the sweet spot is around 1200 to 1500 frame videos. I compare stacks each year and overall the shorter videos provide a slightly better stacked image. Could be atmospheric conditions, could be the position of the planet , could some moon glare who knows , but I’m sticking with the shorter video captured for the moment based on past experience. This obviously may not be the case for everyone but it’s working for me
Cheers

[Stonius]

I did a lot of tests a while back and my 5% stacks tended to turn out better than my 10% stacks (or greater), which otherwise all seemed about the same to me.


In terms of noise, I didn't notice a huge change (doesn't noise reduction fall off after 33+ frames anyway? Or is that only for Deep Sky objects?). Maybe it's time to experiment again if I ever get the opportunity to point my scope skywards again :-(


Cheers


Markus

[DarkArts]

Me too, although it was few years ago with AS2. I haven't repeated experiments with AS3.

With, say, 5000 to 10000 frames (high speed, target dependant), I found the lowest feasible percentage of frames to use was around 4%, below which image quality degraded. Above 10%, there was no appreciable difference.

That's not a hard and fast rule, just a general observation that may be dependant on target, seeing, equipment and user settings.

[Atmos]

In my one and only attempt at planetary imaging, I found that some 80% of frames were over the 50% quality mark according to AS!3, would it still be better to be using a small amount of frames or just go big with 75%?

[Startrek]

In my experience over the 3 years I’ve imaging the planets I would go for 50%
Unless you know the maths behind AS and how the algorithms work in assessing and determining the range of quality then the law of averages suggest start smaller and work upwards
Only my assumption ??

[Tulloch]

One thing to remember about the quality graphs in AS is that it's only relative to the "best" quality frame in the video, not an absolute quality figure. So if you capture 1000000 evenly terrible frames, the quality graph will look great, but the image will look terrible because they are all bad, just equally bad.

Stacking less images might lead to a sharper result, however it will most likely be more noisy and so you wont be able to apply the same amount of sharpening to the final stacked image and so might end up looking worse.

As I mentioned in another thread, with my C9.25 and ASI224MC, my normal process is to image Jupiter for 3 minutes at 150 fps (so 27000 frames) from which I normally use stacks of 3000, 5000 or 7500 frames (seeing dependent) and process which one is better. For Saturn I use 5 minutes at 100 fps (so 30000 frames total), and stack 5000 to 10000 frames. For Mars I use 4 minutes at 250 fps (60000 frames) and stack 3000 - 5000. Neptune and Uranus I image for 5 minutes at 20 fps, normally stack about 50% of them.

I recently tried using WinJupos to derotate longer videos for Jupiter and Saturn, my latest effort was a 15 minute video on Saturn (see image section below). Since I'm on an alt/az mount I was able to use WJ to remove both field rotation and planetary rotation on the video, then stack and debayer in AS!3. Here I was able to stack 10000 frames out of 90000 to get very low noise levels but also able to use high amounts of sharpening without a huge amount of denoise in Registax, so the final result was sharper.

Here is a video showing an interview with Emil about the soon-to-be-released AS!4.
https://www.youtube.com/watch?v=JIjXmRh1DE0

Andrew

[xa-coupe]

After reading a couple of comments I did an experiment. I appreciate all of the comments about quality, and also this is hardly scientific but may be of interest.
I took the same video and ran a stack of best 50% and best 10%. I processed them with the same amount of sharpening and denoising. In my opinion, the 50% is better, but not massively and the difference could well be made up by changing the processing.


I've attached them so people can form their own opinion.

[bojan]

A tad less noisy.