Plane to see

[Tim]

After going through the images for my current project, I found some with trails from aircraft on. There's usually be some satellite trails too but I dont think the sky conditions permitted that. This is a stack of the images with trails on, stretched dramatically to show the fainter ones up.

Trouble is, when I stacked the remaining 128 images, there's still one with a trail on......gotta go back though 'em

Looking at the amount of traffic, its a wonder anybody ever gets a decent pic.

(click to enlarge)

[EA2007]

its those two large fuzzy opnes that haven't moved very far that bother me........they must be pretty close to be that size, trouble is they don't really look like aircraft....could be balloons.

[Psychobilly]

TJ,

What stacking method you using there are a few ways to lose the plane and satellite trails withoust removing the sub...

Billy...

[Tim]

Its usually best to remove the sub altogether. If the image is being stacked in such a way that the trail doesnt show, then it is likely that some of your hard won data is not showing either.

For this image however I deliberately stacked them to emphasise the problem. (entropy)

EA2007, you are right, I hadnt noticed those artifacts, maybe flats will sort them out................

[Psychobilly]

hmmmmmm not to sure about that one...

But i guess Luc Coiffier knows a thing or two...

http://www.cloudynights.com/ubbthreads/showflat.php?Cat=&Board=DSLR&Number=2523450&Forum=f85&Words=&Searchpage=0&Limit=25&Main=2523450&Search=true&where=bodysub&Name=44444&daterange=1&newerval=2&newertype=m&olderval=&oldertype=&bodyprev=

But we'll agree to differ

Billy...

[Tim]

Ok, here I reveal a little of my OCD personality

I nearly always stack all images using all of the methods available, just to see which one works best. If there's a plane trail on it, I havent yet seen it disapppear if it was recorded using one of the faster scopes. Satellite trails come through if the nights are clear enough, but make much less impact. On the C9.25, its usually fairly safe, even at f6.3, but with the F4 OTAs and even the WO72 at f4.8 it struggles to hide the trails. Its a shame there isnt a PS tool to do what the touch up brush does, but do it in a thin straight line!

[MartinB]

That's an amazing image TJ.

You really shouldn't waste those subs. I don't know what the data rejection tools are like in DSS but a sigma reject stack usually works fine in maxim especially if you have 20+ subs. Sometimes you need to tweak the SD a little.

[Tim]

Martin, I dont generally use the sigma reject stack as I have no real clue what the 'sigma' refers to, and therefore would just be altering the figure blind. I dont suppose you happen to know how to get an angle on the figures to use?

Luc Collifer has Kappa-Sigma stacking as being probably the best, but I tend to go with auto adaptive weight oriented stacking as a rule.

These subs add up to 40 mins, in the grand scale of the whole image without these I have 14 HOURS of subs so far, but am still awaiting that elusive really clear night. The flux I am chasing is starting to show through, but processing it is nearly a reverse of usual tactics. I think I may need to double the exposure time to get anything worth bringing out, not sure if its worth it though.

Thanks

[barkis]

Looks like an intergalactic war theatre.

A great looking pic though.

Ron.

[MartinB]

Sigma is just the symbol used for standard deviation. It would be more helpful to call it a rejection method based on standard deviation but sigma reject is shorter! Here's the definition of standard deviation (it isn't from my head, I googled it)

A statistic used as a measure of the dispersion or variation in a distribution, equal to the square root of the arithmetic mean of the squares of the deviations from the arithmetic mean.

Imagine you were wanting to check whether people were anaemic. If you measure the blood count of perfectly healthy people you will find variation around the average which can be plotted to produce a bell shaped/normal distribution curve. If you had a condition making you anaemic your blood count would be way over to the left of the curve. If you were only just becoming anaemic it might not be very far over. So you have to use an arbitary statistical measure for a low blood count being considered abnormal. Typically the value used is 3 standard deviations away from the mean. An SD of 2 would detect a lot of perfectly healthy people, one of 4 would miss a significant number of people with early disease.

If you looked at the value of a particular pixel in a large stack you will get the same normal distribution curve. The chances are that all the stacked pixels will be less than two to three standard deviations from the mean. In this case, rejecting pixels more than 3 SDs from the mean would result in no pixels being rejected and the result is the same as an "average" combine (the best combine for reducing noise when you have clean data). However if a satellite went over or one was affected by a gamma ray burst the value of the pixel would be abnormally high, maybe 5 standard deviations away from the mean. Sigma reject identifies that single pixel and chucks it.

The sigma value or clipping point is normally the number of standard deviations away from the mean you want to select. So, if you have problems with high cloud and a lot of subs you might want to eliminate cloud affected pixels by using a low value (rejects pixels closer in to the mean). If you don't have many subs but have a couple of very bright plane trails to get rid of choose a higher number, maybe 4.

If you only have a few subs, less than 10 then the stats become pretty unreliable.

A good starting point is a value of 2.5 but it's best to try different results to see what you end up with

I guess the software packages add other refinements and may have other things to tweak but that is the essence of the sigma reject.

[themos]

You could just erase everything along a trail (and N pixels either side) and feed the rest onto the stack. Do your own rejection, as it were. One problem with kappa-sigma that I see is that the trail will produce SOME pixels that are JUST not rejected when there is strong signal in the sky. However, they will add to the stack and contribute to the final image.

[Tim]

Thanks Martin, its starting to make sense!

I'll try it tonight and check the results! Thanks for that.

[roundycat]

I'm with Martin on this. I never chuck away subs unless they are spoilt by clouds adding too much noise. I have run many tests on different combination methods and find that SD Mask as interpreted by Maxim DL gives the best result. I looked for minimal noise AND rejection of unwanted data.

The values I use are; No of passes 3, Sigma factor 0.5, Normalisation Linear, Normalisation area 50%.

If you think it might work with a particular trail try cloning it out in PS by using shift-click with the tool. Select your sampling point as normal, start the cloning operation as normal and then go to the other end of the star trail, hold down the shift key and click again. PS will draw a straight line between the two points. Works with any of the painting/drawing tools..

Dennis

[Herakles]

I need a translator !! Weird picture ,War Of The Worlds kinda . Right I am away to try fathom out what you lot are discussing .

I may be some time sadly !

[Roy]

Tj,

Have you seen this, it's shown using AA4 but I'm sure it can be done using other software

http://www.ip4ap.com/Video/Quick_Tips/QT3-Detrail.html

[661-pete]

I was going to try to explain sigma clipping myself, but MartinB has done it far better than I could! I usually use much lower multipliers though: ImagesPlus usually suggests a value of 1.0 (i.e. reject every pixel which falls outside 1*? from the mean). I usually go for about 1.5. This gets rid of most satellites and other noise. Note that the sigma clipping is applied independently to each of the red green and blue channels.

[RobH]

Great pic Tim!

The blue beams are the federation, the red, Klingon

I use SD mask in Maxim. If I have a trail but otherwise the sub is OK I leave it in.

The subs I rejeect are ones with any sort of blurring, or very bad signal to noise ratios.

Cheers

Rob

[Doc]

LOL

I was just about to say it looks like a battle scene from star wars

[MartinB]

Russ Croman has a fantastic sigma reject tool for Maxim DL. You can actually view a stack with just the rejected data, what you get is an image full of trails, gamma ray bursts and hot pixels with no target data. Brill

[Tim]

Thanks for all the tips guys.

It is as I thought, using sigma clipping does get rid of the trails, but also gets rid of the very very faint noise which is the flux I have been trying to pull out. I dont think it is possible to do without really good skies. Adding nine hours of exposure yielded very little more flux. So i'll just concentrate on processing the stack as normal I think. Maybe come back to it after a dark site visit.

Here's a related question then. I have two sessions with a lot of subs each, and different flats for each session.

If I stack them both seperately using sigma clipping, what would then be the best stacking method for the two subs, as really ALL data they contain is useful and should be retained.

Thanks

[roundycat]

To answer your last question here is what I do under similar circumstances. Each sub frame must be calibrated as per normal, ie; dark or bias subtraction, flat fielding then combination. Combine all subs of the same exposure together using SD Mask. Combine the two or more stacks using an average output.

If one of the stacks is noisy bear in mind that the result will be noisier than need be and you may wish to dispose of the noisy stack in the usual place.

Noise can be further reduced by dithering the mount between exposures.

Dennis

[Tim]

Noise can be further reduced by dithering the mount between exposures.

Dennis

Thanks Dennis. I have not heard about this before, what is involved? I usually get told off for dithering

[Psychobilly]

Dithering is very useful... well at least I think it is ... but I am really not sure...maybe it is maybe it isn't who knows...

But seriously...

Dithering is moving the scope between exposures which in helps to smooth out fixed "noise"

Billy...or is it...i used to be undecided but no i am just not sure...

[Tim]

mmmmm, ok, but isn't 'fixed' noise actually signal?

Is there a way to automate that? And im guessing it would only have to be by a few pixels at at time?

Conversely, I found that when using a dslr, with a few hot pixels, the mount was moved between exposures anyway due to flex, not by intent, and the hot pixels showed up as streaks in the image. But I suppose sigma rejection would have gotten rid of that then?

Ron, I hope you are taking all this down!! It's called the vertical learning curve I think

[RobH]

Several programs allow you to dither exposures TJ, certainly Maxim does.

Personally, I find bad guiding works perfectly, my images always move a little bit

Cheers

Rob