deep skystacker

[bejay]

Hi, All

I'm new to the astronomy section, so after reading a few posts, I thought I'd try a few long exposures and feed them into deep skystacker software to see what happened.
So I set my camera on my tripod with my sigy 10-20 at 10mm set ISO to 1600 f4 for 20 secs. pointed the camera at the sky ( I didn't really care what it was pointing at ) set camera for high speed, locked my remote release, and went in and made myself a cup of coffee.
20 minutes or so later released shutter button and found I had recorded 65 frames, the were very orange in colour but that was probably due to stray light, I then fitted the lens cap and did the same again to record the dark frames.
These were then fed into DSS and it did its job and produced a tiff image which wasn't that bad looking. But what am I supposed to do now, I loaded the tiff into CS3 but its just black no stars at all, am I supposed to do something else in DSS. Sorry about the rant, I can't post images as yet as I'm in work

[Adrena1in]

Hi there,

Although I never do any post processing in DSS itself*, I know it always produces very dark TIFF images for me when I load them into PS. But a quick fiddle with the contrast and highlights/shadows brings out a lot of detail.

* Only because I don't know what I'm doing yet, and it always runs extrememly slowly on my old laptop whenever I made any changes to the settings in DSS.

[bejay]

I'm an ar*e,
Only went and loaded the master dark into PS.
Any way below are two images the first is one of the 65 raw frames adjusted with curves and others to get it as best I could, and the other is the autosave from DSS, curves adjustments et c. The red glow in the bottom is I guess light pollution

I think DSS did a pretty good job.

[Adrena1in]

I'm an ar*e,
Only went and loaded the master dark into PS.

Yeah, that will look pretty black! ;) :p

[spit]

you need longer exposures, looking at your photo though, without tracking your not going to a good image

[MintMark]

Bejay your image is OK and very similar to what I get at this stage. The image is dark right out of DSS because it has high bit depth and linear pixel values.

I have been learning to use pixinsight LE for further processing. It can work with 16 bits/channel (preserving dynamic range) and let's you adjust levels and curves. It also has a tool for simulating and subtracting the background gradient. And a couple of noise removal tools too. But I'm sure all that can be done in PS too.

From that image you should be able to get some milky way clouds.

The other problem we have (I say we, because I have a very similar image of the same part of the sky) is the smearing in the corners. I have found that it is worse the more images you stack.

I am beginning to think it is down to deep sky stacker, but I'm not sure. I've attached a map from stellarium with the celestial sphere grid overlaid. All the stars move one square to the right every hour, but as you can see, the direction they move in is different in different parts of the image.

If deep sky stacker is looking for a uniform transformation across the whole image when doing its alignment then it's the corners that will have the most distortion, which is what I'm seeing. I have another picture of the sky including polaris and that one only smears in two corners where the grid is most curved.

Another possible explanation is distortion from the lens, but I don't how to verify that. I've tried exporting images from digital photo professional with lens distortion correction applied, but I still get similar results from DSS.

As I see it we have a few options.
Stack fewer images (less movement, but less noise reduction and fewer faint objects).
Use a narrower field of view (but don't get so much in).
Try some different software (any suggestions anyone?)
Get a tracking mount (the mount corrects for the sky movement leaving DSS to correct for the mount inaccuracies).Anyway, you're heading in the right direction!

[bejay]

Thanks for the reply Mark,
I don't think I will be going full hog into astrophotography, I was just amazed at some of the images posted on this forum, yes I know most of them are taken using very powerfull and expensive equipment, but they compelled me into seeing what could be achieved with standard cameras and lenses, I have taken images of star trails around polaris in the past and was quite pleased with the result, but unfortunatley I live very close to a steel plant and the light polution from there is horrendous, but hey ho if I want better quality images I will have to travel further afield, I was thinking of building a barn door tracker and then using longer lenses and see what happens, at the end of the day its costs only my time !!!!!!

Barrie

[spit]

you can do that, DSS is a fine program- it just cant align the stars with a stationary alt/az tripod- your off in both axis's, if you have no tracking at all- try nearer the NCP and only a few subs

[bejay]

Thanks Spit,

I,m Waiting for a friend to build me a drive module for a stepper motor, when I get it I will build my barn door.

[SteveInNZ]

I did almost exactly the same thing as you and just posted my result in the photo section. I suspect that it is your foreground tree branch that is throwing it off. DSS probably thinks that it is a deepsky object and is trying to align both that and the stars and getting all confused.
Do you have way to crop that out of all of your shots before stacking ?
Oddball idea - I wonder if you could make a dark that included the branch and have DSS subtract it out as it goes ?

Steve.

[Madweasel]

I don't think it's the leaves causing the problem - DSS is cleverer than that. It's more like what MintMark was getting at: in his example he's used a representation of the sky on the screen, whereas the actual problem is the projection of the image by the lens, which causes the same effect. The sky is a dome, but the screen, sensor, print is flat. The lens is designed to portray straight lines as straight lines (rectilinear), but they're not really straight from the point of view of the camera. If you point the camera square-on to a brick wall, the rows of bricks will show as parallel, but really the lines of bricks should get closer together the more you move away from the centre of the image, because they are further away. In most photography that would look horrid though, and would be described as barrel distortion. In this same example, if you now turned the camera to the left and the right, the lines of bricks would change angles to each other (the same as the difference between shooting straight at a building or tipping the lens upwards and making the verticals converge). The same thing is happening as the stars move through the image (caused by the earth's rotation), and that is why stars towards the corners of undriven shots cannot be mapped by DSS.

[MintMark]

Right Madweasel, and I've been doing some more tests. I created three synthetic sky images using stellarium of roughly the same view as some images that I couldn't stack. The images are about 15 minutes apart in time, so half an hour from first to last. In stellarium you can choose the projection it uses to draw the display and I chose "perspective". This keeps straight lines straight (like my lens does) and behaves the same as you move the view around. This test should eliminate sky conditions, camera noise and arbitrary lens distortion from the experiment. It should be what you would see through a perfect rectilinear wide angle lens.

I could not get deep sky stacker to stack those three images at all. I tried star detection thresholds, median filter or not, alignment method (bilinear, bicubic etc), custom rectangles, manually editing the detected stars... I saw plenty of examples similar to the behaviour I see with the real images (doubled stars in the corners and edges with shifts in various directions) but it couldn't register the three images. Many times it couldn't find a transformation at all.

Then I tried the evaluation version of registar with the same synthetic images. It worked first time! It shows pairs of images overlaid and you can see that one has been transformed into a trapezium to make the stars fit. Of course, the trapezium is what you get when you look at a rectangle and apply perspective.

So is the conclusion that deep sky stacker can't detect a perspective transformation between images? It looks like it, but I am surprised. It does explain what I have seen in my images though... the areas that won't stack are those where the stars move towards or away from the centre of the image.

That explains why in my milky way image the corner with polaris was fine (it doesn't move much) and the corner opposite polaris was fine (it moves across the corner but maintains its distance from the centre). The other two corners are basically moving towards and away from the centre and they are smeared. Similarly with the view in this thread... all four corners move compared to the centre and they are all smeared.

So, I'm satisfied that I understand what is happening now and I can avoid this situation with deep sky stacker. It's not happy with wide angle views from a fixed tripod.