Stitching night sky?

[Michael Fontana]

Hi folks

did anybody managed to stitch a night sky, with the moving stars?
I'd like to do that, but am not to sure about the result.

Here's a crop from a single testshot of the real situation, I took recently.

The idea is to get a similar pano like that one, but with the stars.

Meanwhile the zoomify was 3 rows with 7 frames - YCZ Planar 50 mm - each, for the night shot, I tend to make a single row with 5 x Distagon 28, due to the long exposure time of about 15 to 20 minutes at f 2.8!
The shooting will be at the end of year at 2200 m altitude, therefore pretty cold in the required cloudfree night.

I imagine the stitching process not beeing a big problem, as long as I avoid using the stitcher's CP's in the star-aerea of the frames. The scene is orientated towards south, the equator, therefore the star's trace are rather straight lines than circles.

Any insights?

 

[John Angulat]

Hi Michael,
Although I've never stitched a night sky, I've had some pretty good luck a product called "Autostitch". It's a free (imagine that!) piece of s/w developed by post-doctoral engineering students at the University of British Columbia. It's a bit funky to use and you have to remember who designed it (engineers) but I use it for my panorama work.
Here's a link to their site: http://www.cs.ubc.ca/~mbrown/autostitch/autostitch.html
Download the demo version and have a go at it.

 

[Michael Fontana]

Thanks John

I' ve two stitching apps, they work fine; PTGui and Autopano Pro.
The question was rather if its possible to stich night skies.

 

[JohanElzenga]

The problem you will have with stitching night skies is that the stars have moved position between the shots, so the streaks won't align properly. I've never tried it, though, so I don't know how serious this problem will be.

 

[Michael Fontana]

The problem you will have with stitching night skies is that the stars have moved position between the shots, so the streaks won't align properly. I've never tried it, though, so I don't know how serious this problem will be.

Yep, Johan,
that's exactly what I fear as a possible source of problem; well I hope that; lets say the few secs of turning the panohead will not influence that much - compared to the long exposure and the fact, that the stars are moving by 15 deg/hour; therefore 0,0625 deg/15 sec.

But I wonder how the stitcher will react with these lines...

Another thing is that it all happens in about 1 - 1.5 meters of snow; quite 15 min away from the next hut. Avalanches shouldn't be a problem, as they come down during the day, when the snow is getting warm.

 

[Daniel Buck]

The problem you will have with stitching night skies is that the stars have moved position between the shots, so the streaks won't align properly. I've never tried it, though, so I don't know how serious this problem will be.

indeed! I think the best way would be to use multiple cameras and fire them all off at roughly the same time.

 

[Michael Fontana]

indeed! I think the best way would be to use multiple cameras and fire them all off at roughly the same time.

Daniel
another idea, but not really doable:
first organize 5 x tripod, 5 x 1 Ds-2, plus 5 x the same lenses, then carry it about 3 hours up to that place, in deep snow! Plus food for a few days. We will walk with skis and special furs underneath them, for not sliding back:

 

[JohanElzenga]

Yep, Johan,
that's exactly what I fear as a possible source of problem; well I hope that; lets say the few secs of turning the panohead will not influence that much - compared to the long exposure and the fact, that the stars are moving by 15 deg/hour; therefore 0,0625 deg/15 sec.

True. It will make the streaks much longer, but they should align.

 

[Michael Fontana]

Johan
keep your finger crossed.

I might need to "calibrate" the 14 -24 for stitching, as it performs well with the f-stop wide open. I don't think the distagon 28 will match that lens performance, wide open?

The con is another kilo to carry...

 

[erik leeman]

If you want the stars to appear as points you'll need an astronomical tracking device, heavy, bulky and battery dependent. NOT what you'll want to carry up a mountain in the cold. An other option would be to use a Photoshop filter that compresses the elongation exactly along the correct path. Both methods would not be of much use for a stitched panorama though.

I think for this plan to work you need a panohead with VERY accurate stops and a good template for that panohead+camera+lens to use in PTGui. With a setup like this you can (theoretically) stitch your nightsky series without any Control Points linking the shots because the template will take care of their alignment. This technique requires an extremely disciplined and methodical 'operator' and a very expensive panohead! The 360 Precision would come to mind...

Erik Leeman

 

[Michael Fontana]

If you want the stars to appear as points you'll need an astronomical tracking device, heavy, bulky and battery dependent. NOT what you'll want to carry up a mountain in the cold. An other option would be to use a Photoshop filter that compresses the elongation exactly along the correct path. Both methods would not be of much use for a stitched panorama though.

I think for this plan to work you need a panohead with VERY accurate stops and a good template for that panohead+camera+lens to use in PTGui. With a setup like this you can (theoretically) shoot your nightsky series without any Control Points linking the shots because the template will take care of their alignment. This technique requires an extremely disciplined and methodical 'operator' and a very expensive panohead! The 360 Precision would come to mind...

Erik Leeman

Good morning, Erik

sure, for that purpose, a 360 Precision would be nice!
Not having the 360 Precision, but the MA-klickstop (due to a multiple-lens stitch set-up), I hope it will work, even the MA-klickstop isn't as precise, therefore no template:
as the mountain's are parth of the stitch as well, I could set the CP's in PTGui on the snowy mountain and earth only; that's what PTGui does with the day-variant of the same stitch already, automatically.

The first try would be doing autostitching, while deleting the CPs in the sky, as they will distort the other parthes of the images. As last resort, I could stitch manually, with the 4 points- methode/John Houghton.

 

[erik leeman]

There is another technique worth mentioning: the back-to-back method.
The name of it's (Swiss) inventor won't come to mind right now, Matthias Tauchwalder?..., hmm, I'm not sure. Anyway, the trick is to simultaniously shoot opposing shots with two cameras mounted back to back on the panohead. When you tilt the panohead, one camera will look up, while the other one looks down. You can make a template linking those shots, which is clever, because the camera looking down will take shots that will yield good Control Points. And if you can accurately stitch those downward shots you can automatically ALSO stitch the opposing sky-shots. This also works splendidly with cloudless or overcast skies, Matthias uses it for his extremely high res multi-row alpine panos.

Erik Leeman

edit: found correct name
Matthias Taugwalder, http://ivrpa.org/user/496, http://www.concept360.ch/

 

[Michael Fontana]

Erik, that's clever!

I knew Taugwalder's Matterhorn-VR's, but not the back-to-back method.
For the situation in question, it won't help, as I'm not having a 2nd 1 Ds-2......

But I think, as long as I' ve more than 50% of earth/mountain, it should be stichable in PTGui. Still, if one or two frames would have sky/stars only, I could stitch them manually, by chosing the end of the stars trace from frame 5, and the beginnig of the stars line from frame 6; don't you think?

 

[Cem_Usakligil]

Hi Michael,

It may sound silly but why don't you try this out at/near your home on one cloudless night? Just make 2-3 pictures (of the landscape and the sky) using long exposures and see how and if you can stitch them using your trusted tools. If it works, then you are also set for the mountains. If it doesn't, then you'll have prevented a missed opportunity and will be able prepare properly for success. Just a thought....

Cheers,

 

[Michael Fontana]

Thanks, Cem

the dummy is a good idea, and I had thought about it, already.
As for the light polution, I better stay a bit away from the city.

 

[erik leeman]

I could stitch them manually, by chosing the end of the stars trace from frame 5, and the beginnig of the stars line from frame 6

Maybe you could, if you allow for a gap between those trail ends. With the known time interval between shots (which is easy to find in the meta data), that gap could even be calculated I think.

 

[Michael Fontana]

Good support, Erik

actually, that's why this thread is for:
everbody says, that it is impossible to stich, but I want it.

With the help of you all, I might find the best solution, for suceeding.

Beeing well prepared, the turn of the panohead - on top of the MA-klickstopp - shouldn' t take more than 5 - 10 secs.

 

[Valentin Arfire]

hi Michael
I found this
http://www.skywatchertelescope.net/swtinc/product.php?class1=3&class2=306
at Photokina.

but if I get it right, you'd like a bracketed image with stars, so is not necessary a long exposure time.
then use the widest lens you have (extreme fisheye recommended) open the diafragm and try a bracketed exposure;
after that you could try to reduce somehow the noise. Also tufuse might help

regards,

 

[Michael Fontana]

Valentin

I do not want to track, as I want to have the landscape (mountains) to be fix, with the traces of the moving stars. Just like my first testhots - Heather copied that link at the previews post - but with a extended FOV.

The final image should similar to that; with more sky, and at night:

The night- testshot was just a quick and dirt trial, to verify the form of the traces; therefore some noise.

Looking north - at the northern hemisphere - will provide the shape of circle; meanwhile looking south will give the testshots traces - rather long, slightly curved lines.

 

[Valentin Arfire]

star sky

hello

let's say for a given time - that remains to be discovered - the stars may be approximated as single pixel dots.
after the given time you'll have two pixel instead of one.
what if you take several series of the same picture with the time.
the stars will apparently move but I suppose each to other will be at the same position, so by overlapping the stars you'll obtain the stronger contrast sky with less noise.
then just choose the image you like the best with the mountain and use only that mountain in the final.

I suppose besides the cold it raises some other photoshop problems also...

my idea

regards,

 

[Michael Fontana]

hello

what if you take several series of the same picture with the time.
the stars will apparently move but I suppose each to other will be at the same position, so by overlapping the stars you'll obtain the stronger contrast sky with less noise.
then just choose the image you like the best with the mountain and use only that mountain in the final.

I suppose besides the cold it raises some other photoshop problems also...

my idea

regards,

yep, the cold might be a problem, one year there was minus 17 degs Celsius, at that place, at 5 pm... so there's not really a obsession to make it through the night... ;-)

if it will be really cold, the next stitch might be a better scenario, as it's taken not far from the cabine; warming up is possible.

The image is taken with the 28 mm/FF. By using a 1.7/50 Planar and 2 rows with 4 frames each, there should be a big print possible, whithout bracket shots.....
to be precise, the stich will have a wider FOV.

A problem could became the frame at the right upper corner, as there's not a big overlap in the landscape, to set the CP's....

BTW: does anybody knows how long the battery (NP-E3) will last in these cold conditions? Any °keeping warm° -advises? I ordered a 4th delkin yesterday...

 

[Michael Fontana]

The star traces seems to stitch well, as long as you' ve some CP's as reference points.

I found Stellarium, a programm which shows the star-constellations for a customized day, time, orientation and place, pretty handy for that project.

I filmed it from screen to make the movement of the stars, becoming traces in the long exposures visible.

So this is the sky, looking northeast, the 26th of december, from 8 to 12, for the mountain in post 21. Only the center, upper parth will be visible in the stitch, as the mountain hides the lower parth.

 

[Michael Fontana]

The first wheater forecast are available..... minus 20 deg Celsius - at night.. - that's about -5. °F :-(

 

[Michael Fontana]

My best wishes to you all for 2009!

back again, as I 've a assignement this night at 4 am; the opening of a new custom facility between france and switzerland....

yep; have been at the alps, with lots of snow and some quite windy days .... still it was possible to take some night photos and stitches but I need more time to get it worked out. Meanwhile, just as single shot, taken the first of january 2009 and showing the little village:

I stayed about a week in the hut with the 2 candle-lightend windows - the first days nobody was in the little village, and the hut has been rather cold, when arriving. Later, some friends arrived, and we spend a few days, and new year's eve together, which is great in that environnement.

As it has been pretty windy, one of the problems with stitches has been the stability of the tripod/cam in the soft snow; therefore I had to set up the tripod during day, while shoveling the soft snow away, and °glueing° the tripod with water on the lower, harder snow surface.

 

[nicolas claris]

Hi Michael

SIMPLY BEAUTIFULL

A keeper!

 

[Valentin Arfire]

beautiful

an inspiration in your picture.

After the vacation I think everybody is waiting for more

regards,

 

[Michael Fontana]

Valentine, I had a job to get done first...

so here's the first nightstitch:

3 x 30 min, distagon 28, at f = 4, the stars aren't stitched together - they don't show 90 min but 3 x 30 min - but the landscape.

At the right side, there's Orion, with the 3 paralell lines.

Orion in a screenie at 100%:

Some other long exposures showed some nasty lines. It' s a shift in color as well as in luminance! Does someone know it's reason?

 

[Klaus Esser]

Thanks John

I' ve two stitching apps, they work fine; PTGui and Autopano Pro.
The question was rather if its possible to stich night skies.

Hey Michael!

The upcoming APP versions read the script of a panohead and place the logical order first instead of seeking CPs. That deals with blue skies also as with night skies, water or walls and so on.

Nevertheless you can move/place pictures or sets of pictures manually in APP 1.4.2.

best, Klaus

P.S.: i have a motorized panohead - the Merlin/Orion - which in fact is an astronomy tracking head.
You can set it to earth rotation and have the stars as points instead of stripes.
These shot i would combine with a longtime-exposure of the mountains.
This way you can have both correct displayed.

(i didn´t read all postings - maybe this was suggested already)

 

[Michael Fontana]

Thanks Klaus

I' m fine with my equipement; just don't expect me to carry that Merlin high up to the alps ;-)

Nobody having a idea about the source of these nasty lines in the last picture in post 27, and how to get rid of it? I could stitch another night pano but would have to resolve the line-problems first.

That problem doesn't shows up at all with exposures like 8 minute or shorter and in day shots.

 

[Klaus Esser]

That problem doesn't shows up at all with exposures like 8 minute or shorter and in day shots.

Seems to be related to the 30min exposure-time. That´s very long for digital shooting using a normal camera.

Besides:
1) the Merlin is lighter than a Manfrotto SPH
2) you save a lot of time at stitching using a scripted shooting.
3) i wouldn´t be high up to the alps anyway . . )

best, Klaus

P.S.: when it´s very cold or/and a need for constant power over hours i use external batteries in an isolated box with some small hand-warmers (you put them in boiling water for some minutes - they stay warm for hours when put in such an isolated box.)