unguided / summing lights

[funkysandman]

Can I sum images before stacking to increase exposure time?

Let say I take 20 lights at 60sec & ISO 800 and things look very faint - could I group the set into pairs, sum them, then stack?

..so would it be like having 10 lights at 120sec & ISO800?

btw, does "subs" mean the same as "lights"?

[wimvb]

That would be a nice trick, but unfortunately this doesn't work as you'll also add the noise by the same amount. Nothing would be gained.

In order to get more detail you take many sub-frames (light frames) to average out the noise from them. Then you stretch the image to push detail from the lower data range to higher range.

E.g. if most of your data in one image is in the range 0 ... 10, while black to white is a range of 0 ... 100, then you can push the data to, say, 0 ... 50 by stretching.

Adding the same frame over and over, you'd still have the data in the lower  10% of all possible values, no matter how many sub-frames you'd add.

light frames are only one kind of sub-frames, other are the calibration frames: dark frames, bias frames and flat frames. There are several good articles describing the roles these play in image processing. Google and learn ...

Good luck

[funkysandman]

That would be a nice trick, but unfortunately this doesn't work as you'll also add the noise by the same amount. Nothing would be gained.

In order to get more detail you take many sub-frames (light frames) to average out the noise from them. Then you stretch the image to push detail from the lower data range to higher range.

E.g. if most of your data in one image is in the range 0 ... 10, while black to white is a range of 0 ... 100, then you can push the data to, say, 0 ... 50 by stretching.

Adding the same frame over and over, you'd still have the data in the lower  10% of all possible values, no matter how many sub-frames you'd add.

light frames are only one kind of sub-frames, other are the calibration frames: dark frames, bias frames and flat frames. There are several good articles describing the roles these play in image processing. Google and learn ...

Good luck

ok, subs refers to all dark/flats/bias/lights...got it

[dph1nm]

Can I sum images before stacking to increase exposure time?

Let say I take 20 lights at 60sec & ISO 800 and things look very faint - could I group the set into pairs, sum them, then stack?

Yes, but why? Stacking IS summing (usually plus some averaging at the end), so you would just be doing the same thing as stacking all 20 in the first place.

NIgelM

[Astrosurf]

When people say 'subs' casually they usually mean lights.

Alexxx

[funkysandman]

Yes, but why? Stacking IS summing (usually plus some averaging at the end), so you would just be doing the same thing as stacking all 20 in the first place.

NIgelM

how is summing stacking?  If I take two pictures and sum them, it's not the same as averaging them.  Constant light would be amplified, but noise not as much, no?  The chance for two random noise signals to be on the same pixel is less likely than stretching a single image which would guarantee an equal boost in noise.

[funkysandman]

That would be a nice trick, but unfortunately this doesn't work as you'll also add the noise by the same amount. Nothing would be gained.

In order to get more detail you take many sub-frames (light frames) to average out the noise from them. Then you stretch the image to push detail from the lower data range to higher range.

E.g. if most of your data in one image is in the range 0 ... 10, while black to white is a range of 0 ... 100, then you can push the data to, say, 0 ... 50 by stretching.

Adding the same frame over and over, you'd still have the data in the lower  10% of all possible values, no matter how many sub-frames you'd add.

this is incorrect, "In fact, adding together any number of images will multiply the signal by that number while only increasing the noise by the square root of that number"

[Pompey Monkey]

this is incorrect, "In fact, adding together any number of images will multiply the signal by that number while only increasing the noise by the square root of that number"

This statement is dependent on the the noise source being coherent or random.

If you are summing the noise from identical images, then it is coherent. In which case it will sum at the same rate as the (coherent) signal. The S/N ratio will therefore not change and you will gain nothing. Nada. Nowt. etc.

Stacking works because the signal is (effectively) coherent, but the noise is random.

This noise is usually through quantum effects in the electronics or variations in the circuitry introducing patterns.

Stacking, as I know it, works in two ways (there may be more):

Statistical rejection of data that is not consistent between images (quantum noise, satellites etc.);

Through dithering, where the images are deliberately pixel shifted by design so that each image has a different noise pattern which effectively de-coheres the noise, an the statistical rejection of data can work.

[funkysandman]

This statement is dependent on the the noise source being coherent or random.

If you are summing the noise from identical images, then it is coherent. In which case it will sum at the same rate as the (coherent) signal. The S/N ratio will therefore not change and you will gain nothing. Nada. Nowt. etc.

Although the images are not truly identical, two independent shots

[Pompey Monkey]

Although the images are not truly identical, two independent shots

In which case it is stacking. Apologies if I am being a little thick, but what are you really trying to do?

Stacking pairs, then stacking the pairs of pairs, gives no benefit over conventional stacking.

[funkysandman]

In which case it is stacking. Apologies if I am being a little thick, but what are you really trying to do?

Stacking pairs, then stacking the pairs of pairs, gives no benefit over conventional stacking.

I was working in photoshop and you can take a stack and choose the stackmode: mean, median, summation, entropy, maximum, minimum, range, standard deviation etc...

In this case, take pairs - stack them using summation effectively halving the number of pics, then stack the results using median stackmode

[Pompey Monkey]

I was working in photoshop and you can take a stack and choose the stackmode: mean, median, summation, entropy, maximum, minimum, range, standard deviation etc...

In this case, take pairs - stack them using summation effectively halving the number of pics, then stack the results using median stackmode

Well, in my limited (but I hope valid) experience, you cannot extract what is not there in the first place.

You can't polish the proverbial ****

(Despite Mythbusters doing the exact opposite!  

[funkysandman]

Well, in my limited (but I hope valid) experience, you cannot extract what is not there in the first place.

You can't polish the proverbial ****

(Despite Mythbusters doing the exact opposite!

the idea here is to reduce the need for guiding, by taking 2 images instead of one with same overall exposure time... I will it give it a go with a set of lights I have and play around

[ollypenrice]

No free lunches in this game. The best way to combine sub exposures is by using the algorithm best suited to the number of subs being combined. Larger numbers benefit from sigma rejection to exclude outlying pixel values (arising from things like sat trails and incoming aliens...)

I disagree with Wim on the minor point of the meaning of 'sub exposures.' They are called 'sub' exposures because they are light frames which do not constitute the entire light exposure. Darks, flats, bias etc are not sub exposures, they are calibration files.

Olly

[wimvb]

I think that some of the confusion in this discussion is due to me misinterpreting the original post. Clearly, adding THE SAME FRAME (as I stated in my original reply) to itself can never work.

Olly, thanks for clearing up my confusing gibberish regarding subs.

If I haven't completely missed the point of all my taking exposures and preprocessing them, here is what I have concluded so far:

Bias frames are used for creating a master bias file

Dark frames are used for making a master dark file with the aid of the master bias file

Flat frames are taken for creating a master flat file also with the aid of the master bias file.

Light frames are taken for creating a master light file with the aid of the other three master files.

As such, all the frames we collect with our cameras are basically sub-frames that are used in the creation of the respective master files. What we usually refer to as subs are the light frames that hold the valuable information we're after.

When we create the master files, we combine the frames that go into it, by applying various statistical tools such as averaging, taking median values or whatever you can come up with that will give smooth results.

Please correct me if I'm wrong, in which case I will definitely shut up until I've reread my astrophotography 101.0 ;-)

Clear skies

sent from my mobile device

[dph1nm]

how is summing stacking?  If I take two pictures and sum them, it's not the same as averaging them.

Yes it is. You get the same s/n improvement from summing as from averaging (which is just summing then dividing all the numbers by 2!).

NigelM

[ollypenrice]

I think that some of the confusion in this discussion is due to me misinterpreting the original post. Clearly, adding THE SAME FRAME (as I stated in my original reply) to itself can never work.

Olly, thanks for clearing up my confusing gibberish regarding subs.

If I haven't completely missed the point of all my taking exposures and preprocessing them, here is what I have concluded so far:

Bias frames are used for creating a master bias file

Dark frames are used for making a master dark file with the aid of the master bias file

Flat frames are taken for creating a master flat file also with the aid of the master bias file.

Light frames are taken for creating a master light file with the aid of the other three master files.

As such, all the frames we collect with our cameras are basically sub-frames that are used in the creation of the respective master files. What we usually refer to as subs are the light frames that hold the valuable information we're after.

When we create the master files, we combine the frames that go into it, by applying various statistical tools such as averaging, taking median values or whatever you can come up with that will give smooth results.

Please correct me if I'm wrong, in which case I will definitely shut up until I've reread my astrophotography 101.0 ;-)

Clear skies

sent from my mobile device

If not using a set point cooled camera. If you are using such a camera then the dark doesn't need the help of a bias. The bias is in the dark anyway. The bias can, though, be used to scale darks not at the same temperature as the lights. At least, such is my understanding but I'm no guru on this. Just be careful not to double subtract the bias.

Olly

[funkysandman]

Yes it is. You get the same s/n improvement from summing as from averaging (which is just summing then dividing all the numbers by 2!).

NigelM

You're right...s/n is the same!  I can't polish the proverbial ****