Night sky pics – tips wanted

[chupathingie]

There are several ways to stack, but the simplest method is to is to stack multiple exposures on top of each other in layers. The bottom layer gets 100% opacity, second gets 50%, third gets 33%, 4th gets 25% etc. This gives a straight average of image data, which reinforces the actual data while minimizing noise. Stacking 30+ images knocks the noise floor *WAY* down, and allows you to aggressively stretch levels (always do this in 16bits/channel) to increase the contrast of the faint fuzzies in the sky. This even works with 8-bit channel images, stacking can give you a 16-bit image that can be stretched far more than if averaged into an 8-bit image. It’s not as good as starting with 16-bit RAW-to-tiff, but it’s an improvement over the base 8-bit.

[SilverStag]

There are several ways to stack, but the simplest method is to is to stack multiple exposures on top of each other in layers. The bottom layer gets 100% opacity, second gets 50%, third gets 33%, 4th gets 25% etc. This gives a straight average of image data, which reinforces the actual data while minimizing noise. Stacking 30+ images knocks the noise floor *WAY* down, and allows you to aggressively stretch levels (always do this in 16bits/channel) to increase the contrast of the faint fuzzies in the sky. This even works with 8-bit channel images, stacking can give you a 16-bit image that can be stretched far more than if averaged into an 8-bit image. It’s not as good as starting with 16-bit RAW-to-tiff, but it’s an improvement over the base 8-bit.

So you take, say, 32 images at the same exposure, and arrange them in layers in Your Favorite Image Hacking Program? Am I understanding this correctly?

[chupathingie]

yup. Or you can download one of the free automated tools built for the purpose, or purchase one of the commercial ones. Personally, I’ve had better luck with the manual stacking (as tedious as it is). I’ve not stacked anything in ages, tho… I really need a decent mount to extend my exposure times.

[ravnostic]

There are several ways to stack, but the simplest method is to is to stack multiple exposures on top of each other in layers. The bottom layer gets 100% opacity, second gets 50%, third gets 33%, 4th gets 25% etc. This gives a straight average of image data, which reinforces the actual data while minimizing noise. Stacking 30+ images knocks the noise floor *WAY* down, and allows you to aggressively stretch levels (always do this in 16bits/channel) to increase the contrast of the faint fuzzies in the sky. This even works with 8-bit channel images, stacking can give you a 16-bit image that can be stretched far more than if averaged into an 8-bit image. It’s not as good as starting with 16-bit RAW-to-tiff, but it’s an improvement over the base 8-bit.

So you take, say, 32 images at the same exposure, and arrange them in layers in Your Favorite Image Hacking Program? Am I understanding this correctly?

32 is overkill. Take 32 of the same exposure, but let the stacking software use the best 9 of them. I was reading in this month’s Astronomy magazine that the noise reduction ‘rules’ offer diminished returns for more pictures in the stack. 4 pictures is 1/2 the noise, 9 is 1/3, 16 1/4, 25 1/5, 36 1/6th, etch. The article writer suggests 9.

Of course, I’ve never figured out how to make registax do a good job of it myself, so I’m just quoting an expert. My experience in this area is lacking.

[chupathingie]

I think the author of that article is lacking… 32 is definitely not overkill. Those diminishing returns offer a significant increase in the amount of detail you can tease out of an image.With only 9 subs, you can’t stretch levels very far. You don’t start getting a smooth, somewhat noise-free background until you hit 20+ subs. Seriously, the more the better. More data= cleaner images.

This guy knows what he’s talking about… lots of good info for those wanting to get to grips with what it takes to get good data for images:
http://www.saratogaskies.com/articles/cookbook/index.html

Also, if you’re going to stack, take the time to shoot off 1/2 dozen darks with the lens cap on of the same duration and (important!) at the same temperature. I shoot darks in the middle of my shoots. These can be averaged and used to reduce noise and hot pixels before you start to stack.

This was done without darks, shot as JPEG (this was my 1st attempt, I barely had a clue). In other words, done wrong from a noise reduction standpoint. 36 exposures and shot from the middle of Dallas/Fort Worth. Horrible light pollution.30 second exposures. I should have shot in RAW (duh!) and taken the time for darks. A dozen well-collected subs will just start to get you to where you can stretch the contrast to the point of making the wispy details come out without making all that noise come out as below.

Have a look here:
http://ap.x-wing.ca/gallery/v/Nebulae/NGC7635_67x600s_PI_A1_A1.jpg.html?g2_imageViewsIndex=2
Look at the stats… 67 10-minute subs. Obviously done over multiple sessions (no, not one of mine…someday, maybe…).

[ravnostic]

chupaWow. Amazing detail. As I said, I ain’t no expert. Shame on me for believing what I read without looking further into the matter.

[orionid]

I haven’t put into practice yet, but part of what I’ve read recommends doing darks before, middle, and after shooting to catch all the pixels, including the ones that “heat up” durring the shooting.

Also, for people like me in humid but teperate areas, winter shooting is considered best for three reasons 1) natural cold keeps sensor cool in leiu of expensive cooling rigs. 2) below freezing keeps atmosphere humidity and therefore distortion down. 3) no fear of dew on lens in the post-midnight temp drops. The handful of shooting I have done, while craptacular in output, was all in the winter and the dew/humid rules seem to be accurate.

[chupathingie]

I can see the sense in doing darks before/during/after. Getting clean images of low-light targets is time consuming with a DSLR. It’s a good thing that celestial objects don’t move much.

I’ll second the 3 reasons for winter shooting. The skies in the Adirondacks were amazing in winter. Too friggin cold to sit by the scope for hours, tho. Below 0F was common enough. Brrrr.

[orionid]

So here’s a stacking question.

When manually stacking through PS or similar, how do you accurately rotate the image for best alignment accounting for sidereal motion?

[ravnostic]

That’s actually a question I can address, Orionid, but not really answer. The axis point from which to rotate is likely out-of-frame, technically. Take Orion as an example, it’s far from the north pole, so the axis point would be some 45+ degrees out of view. There’s probably a formula out there that would tell you how far to tweek the image, but it’s not as easy as ’15 degrees for every hour’. The amount would vary depending on the centering of the image and the distance to the north pole.

Thus far, my manual field rotation attempts to compensate have been far lacking. I’ve tried triangulation to two stars in the FOV, but haven’t figured out how to get it right as of yet (you’d think it’s as easy as x^2 +y^2=z^2, but it hasn’t proven to be–despite shooting at the same focal length, the images never quite work out correctly across the image when I’ve done so–I can get two-star alignment, but other stars somehow do not line up when I do so. If someone could solve this quandry for me, I’d be most appreciative.)

[orionid]

Yeah, that’s what I was figuring thus far. I think your two-star alignment is dropping off due to lens distortion, the same effect that comes into play when rotating your camera for a panoramic. The only way to make sure it’s uniform across all the exposures is to have the center of your front element at the center of rotation.

[ravnostic]

The only way to make sure it’s uniform across all the exposures is to have the center of your front element at the center of rotation.

Through the scope, the center will be whatever the point the telescope is centered upon. But this hasn’t helped any in the alignment front. Something about the rotation algorithms seems to throw things off, or perhaps some other optical technicality that I haven’t mastered mathematically to compensate for (and likely won’t, which is unfortunate.)

In *brighter* news, there is Comet Hartley II this month, and it’ll be around Cassiopeia and Perseus this month; a good 30 second exposure might capture a nice view of it, given the field of rotation that near the North pole is minimal over so short a span.

[chupathingie]

I think your two-star alignment is dropping off due to lens distortion

That’s the heart of it, I’m thinking. I’ve been wanting something that can “de-pincushion” images for ages, so I can stack short exposures from a stationary camera. So far I’ve had no luck and have not revisited the idea in some years.

[sleeping]

DXO Optics can do stuff like that (if they have created modules for your camera/lens combination): http://www.dxo.com/us/photo/dxo_optics_pro/optics_geometry_corrections

[caradoc]

Capture NX can de-fish for most Nikon lenses.

[Author]

Posts