Dithering with DSLR astro shots

[LooseFur]

Hi all,

I use a nikon d810 (high megapixel count) and learnt about dithering after a night shoot (of course!).

I understand it means to move the camera/telescope a few pixels randomly in the hopes to reduce noise (pattern/random) when stacking. However, for my night shoot recently I didn't move the camera at all (I didn't know about dithering). 
So, my question is, can you 'dither', after the fact - i.e. does cropping the images you have taken (I only did 10 lights) in random directions simulate the dithering effect? So long as all images are the same size? Or is this idea flawed? When you have 36mp per file, a small amount of cropping is insignificant if it means I can reduce some of the noise.

 Looking forward to your replies.

[MarkAR]

12 minutes ago, LooseFur said:

can you 'dither', after the fact

No, dithering is the mechanical movement of the mount.

[R26 oldtimer]

No because the noise still appears at the same pixels relative to the signal.

[LooseFur]

hmm interesting, but what I was proposing is to crop the images, in a manner which shaves off different sides each time, so the noise would not be relative to the stars in each image. So, for example for image A I crop it losing pixels from the bottom and right sides. For image B I crop it to lose pixels from the left and top sides and so on for various permutations. Would that work?

[Ouroboros]

Nope.  You have to move the mount on each shot so that the stars fall on a different part of the imaging chip. 

[Seelive]

The idea of dithering is to ensure that when the stars are aligned during stacking they are located at different camera pixels in each image.  Cropping the images edges will not achieve that since stacking will just realign all the stars and hence the camera pixels again.

If your tracking is slightly off so that each of the images are slightly offset from each other prior to stacking (the usual case), then that effectively has the same effect as dithering, but since the offset between images is not random, it will be slightly less effective.

[Paul M]

You'll need a goto style mount that can take guiding inputs from a computer/controller and software like APT to opperate the camera and move the mount to create the dither in sync.

The idea being that hot/dead pixels and dust bunnies etc remain in a fixed point in the image frame but the target is moved a few pixels on the sensor before the strat of each new exposure, that movement is "dithering". So when aligned and stacked the stars and things all realign and their data are combined but the unwanted artifacts are not stacked and get filtered out by the stacking sofware, the artifacts don't get combined.

So there has to be some method of getting the wanted data onto a slighly different part of the sensor between frames. It can't be reproduced after the fact.

[LooseFur]

15 minutes ago, Seelive said:

The idea of dithering is to ensure that when the stars are aligned during stacking they are located at different camera pixels in each image.  Cropping the images edges will not achieve that since stacking will just realign all the stars and hence the camera pixels again.

If your tracking is slightly off so that each of the images are slightly offset from each other prior to stacking (the usual case), then that effectively has the same effect as dithering, but since the offset between images is not random, it will be slightly less effective.

Thanks everyone for the replies, you're all very helpful.
I am actually not an astrophotographer, per se - I am a landscape photographer so I don't have motorised/equatorial mounts, and I am shooting using a 35mm Ziess lens (incredible optics)- as such there was no tracking. SO, the stars would be on different pixels for each frame due to earth rotation. If I wanted to dither, I'd have to do it by hand. Pretty tricky to move only a few pixels on a ball head tripod but I have a steady hand, I guess.

Seelive, what you said makes sense, but maybe for tracking mounts? Bear in mind I just use a 'regular' tripod. 

[Seelive]

... Bear in mind I just use a 'regular' tripod

Well that will certainly provide the necessary movement between the images.  Leaving a random interval between images and occasionally re-centring the target will give you the same effect.

[Ouroboros]

Taking a series of darks is the traditional alternative to dithering. That’s where you put the cap on the camera lens and take a series of photographs at the same exposure settings as the lights. These are best taken when the camera is at the same temperature as when the lights were taken. Since you’re on a static tripod I’m guessing your exposures were short. So taking maybe 20 dark exposures is not that onerous. These darks are then averaged together and subtracted from each light image before these are registered, aligned and stacked. 

[Paul M]

31 minutes ago, Seelive said:

... Bear in mind I just use a 'regular' tripod

Well that will certainly provide the necessary movement between the images.  Leaving a random interval between images and occasionally re-centring the target will give you the same effect.

Too simple...

No one likes simple solutions... 

[LooseFur]

1 hour ago, Ouroboros said:

Taking a series of darks is the traditional alternative to dithering. That’s where you put the cap on the camera lens and take a series of photographs at the same exposure settings as the lights. These are best taken when the camera is at the same temperature as when the lights were taken. Since you’re on a static tripod I’m guessing your exposures were short. So taking maybe 20 dark exposures is not that onerous. These darks are then averaged together and subtracted from each light image before these are registered, aligned and stacked. 

Thanks again guys, you are all extremely helpful!
I took only 1 dark at the time, but I do use LENR. In hindsight I now know I needed more darks probably but interestingly enough, when I stack  the 9 light frames with the 1 dark I get a worse result than when I don't (See my other thread here). I also have 40 odd bias frames taken at same ISO/aperture but at 1/8000 shutter speed. The lights were taken at f/5.6, ISO 6400, 4 seconds. f/5.6 is the optimal edge to edge sharpness of the zeiss (Milvus f/1.4) glass (coma aberration at wider apertures, bad star trailing at exposure > 4 seconds). I'm a pretty technical landscape photographer, I focus stack all my landscape shots but this astro stacking is quite new to me!

As for noise, I am cutting it down a bit but as a focus stacking landscape photographer, pixel detail is a big thing for me! So I'd like to improve it.
This snip is from 100% (pixel level) in the top left corner. Any hints on reducing it further would be most appreciated! Bear in mind this is 36mp.

P.S. I am also pulling the image up about 3 stops in Lightroom

Edited July 22, 2020 by LooseFur

[Seelive]

Yes, just using 1 dark frame will be worse than using none.  You really need to stack several dark frames (I always try for at least 20 to 30) for them not to effect the noise in the final image.  The more lights you stack, the lower the final image noise will be (in theory reduced by the square root of the number of lights) so just keep adding to them (up to a point anyway, the law of diminishing returns comes into play!).  

Is LENR the in-camera noise reduction?  If so, you could test the results with and without using it but I think the general census of opinion is to turn it off, it can give more problems than it solves, and just use darks.

The star images certainly look sharp but I think I would be tempted to open the lens up by a least a stop or two. With a F1.4 lens, I would try F2.8, you will get considerably much more detail in the images for the same exposure time at the expense of some distortion in the corners, but you could always crop the image. 

[LooseFur]

14 minutes ago, Seelive said:

Yes, just using 1 dark frame will be worse than using none.  You really need to stack several dark frames (I always try for at least 20 to 30) for them not to effect the noise in the final image.  The more lights you stack, the lower the final image noise will be (in theory reduced by the square root of the number of lights) so just keep adding to them (up to a point anyway, the law of diminishing returns comes into play!).  

Is LENR the in-camera noise reduction?  If so, you could test the results with and without using it but I think the general census of opinion is to turn it off, it can give more problems than it solves, and just use darks.

The star images certainly look sharp but I think I would be tempted to open the lens up by a least a stop or two. With a F1.4 lens, I would try F2.8, you will get considerably much more detail in the images for the same exposure time at the expense of some distortion in the corners, but you could always crop the image. 

Thanks for the reply. Yep, LENR is in camera noise reduction. I will turn this off next time, at least I'd be able to take 2X more images now! I will take more lights/darks instead.

Yeah I tried a few test shots at f/2.8 and f/4 but I really wasn't pleased with the corner performance. I think I would have to crop a lot. This is with a £2k 35mm lens. I am (and so is the nikon d810) very demanding! I tried a few other lenses too (I have the Sigma art 35mm f/1.4 too, it's not an even a comparison if I am honest).
I made a conscious decision to trade noise (use a smaller aperture and therefore higher ISO) for edge to edge detail. Maybe cropping might have been better, but to be honest I can pull the exposure up quite a lot in lightroom with the D810 - it has a good dynamic range latitude. 

Thanks for the very knowledgeable replies everyone, still finding my way a bit and I don't get many clear nights where I live.