The SMC with canon 350D and 300 sigma Zoom

[anthony.tony]

Hello I shot the SMc with Canon 350D with 300 sigma Zoom. I experimented with processining . I used Neat image on one.I copied the same image 70 times then loaded it in deepsky stacker with neat image file and stacked them. What hapepens when you copy the same iamge say 70 times and stack .I used CS3 and saturatiopn Increase then gaussian blur to a Lumanance copy .reguards tony. the exposure was ISo-800 about 210 secoNDS. Tony. What would happen if you strecthed out hte copied frames in Cs3 With levels and curves etc so you are changing each copied frame then stack them Tony.

[turbo_pascale (Rob)]

Stacking the same image many times has absolutely no effect.
The nature of stacking to to remove the noise and randomness in the images.

You will get the exact same result doing your processing on 1 image.

On the other hand, if you went to the effort of doing flats, darks, bias frames etc, you could enhance your result.

Turbo

[Peter Ward]

This is interesting as it may show a mis-understood element others may have about signal, noise and stacking.

Stacking only works if you are adding signal to an image at a rate faster than you are adding any noise. As a result using the same image data its very much a case of if you divide the same number by itself, you get one, no matter how big the number.

Similarly if the image has a high noise level, through say a short exposure of say a DSLR of 30 seconds or less, you are still not building up signal very quickly. Akin to dividing a number by a little less than itself...sure more than one, but not a lot more.

Stacking works very well when the signal swamps the noise (eg, 20 minute rather than 20 second exposures).

As for the Sigma 300mm F2.8. It's not bad. but needs to be stopped to about F4 and you also need to drop a UV+IR block filter in the filter bay.

Hope that helps

Peter

[skwinty (Steve)]

Originally Posted by Peter Ward

This is interesting as it may show a mis-understood element others may have about signal, noise and stacking.

Stacking only works if you are adding signal to an image at a rate faster than you are adding any noise. As a result using the same image data its very much a case of if you divide the same number by itself, you get one, no matter how big the number.

Similarly if the image has a high noise level, through say a short exposure of say a DSLR of 30 seconds or less, you are still not building up signal very quickly. Akin to dividing a number by a little less than itself...sure more than one, but not a lot more.

Stacking works very well when the signal swamps the noise (eg, 20 minute rather than 20 second exposures).

As for the Sigma 300mm F2.8. It's not bad. but needs to be stopped to about F4 and you also need to drop a UV+IR block filter in the filter bay.

Hope that helps

Peter

To put this into perspective.
1 The signal to noise ratio for a number of summed images is equal to the sqare root of the number of images multiplied by the mean photon flux.
2 If you compare the snr of a single image to the snr of a summed image, then the the summing improves the SNR by a factor of the square root of the number of images.
3 So if you want a four times improvement over the single image then you need to sum 16 images.

I think this would apply regardless of whether the exposure times are 30 seconds or 30 hours.
Regards
Steve

[Peter Ward]

Originally Posted by skwinty

1 The signal to noise ratio for a number of summed images is equal to the sqare root of the number of images multiplied by the mean photon flux.

Regards
Steve

But, there's the rub, the mean flux can vary enormously.

With a bright lunar image the S/N is so high as to make the amplitude of the stacked signal + noise almost 1:1 compared to an exposure equal to the total stack time....

All a bit moot really.

Take lots of frames.

Make sure they have more than just noise.

[skwinty (Steve)]

Originally Posted by Peter Ward

But, there's the rub, the mean flux can vary enormously.

With a bright lunar image the S/N is so high as to make the amplitude of the stacked signal + noise almost 1:1 compared to an exposure equal to the total stack time....

All a bit moot really.

Take lots of frames.

Make sure they have more than just noise.

I disagree.
I took a moon pic at 250th/second and did not calibrate or use noise reduction and the image showed no noise.
Now I could have taken 20 pics and stacked them and still improved the SNR
noise = square root of mean photon flux.
This is disregarding camera induced noise which can be calibrated out.
Now I agree that the flux can vary enormously, but then so does the noise.
If you point a dslr at the sky from a light polluted suburban sky and expose for 30 seconds you sure will get more signal than noise .Unless you using a pin hole camera of course then you probably wouldnt get any signal.
Great discussing these points with you
Regards steve

[Peter Ward]

Nup.

(the cool thing is I am operating my scope remotely while I have this chat )

You can not dis-associate the signal from the noise.

And there is so much noise.

Dark Current,
Generation-recombination,
diffusion, surface, back & substrate dark current,
shot,
MPP,
spikes,
spurious charge,
transfer, residual bulk & surface,
luminescesce (pixel, clocking and diode),
cosmic rays,
excess charge,
light leak,
preamp, A
DC quantizing,
clock jitter,
emf,
decoupling,
shielding &
crosstalk......

Originally Posted by skwinty

noise = square root of mean photon flux.
This is disregarding camera induced noise which can be calibrated out.

It's simple enough. Lets say we barely open the shutter. Not much signal
but stack the frames anyway. The noise will increase far faster than the signal, and we get? Noise. The signal is buried

Then we take a deep exposure. Buckets of signal. We stack the frames.
And we get? Signal. The noise is buried.

Sure the square root association is there, but not in isolation.

[skwinty (Steve)]

Now thats way too cool. I am boiling in the heat here in Cape Town and am not at work with ADSL so having to resort to 56k dial up. So needless to say I am offline while I compose my message.
Perhaps in my next life.
Firstly lets look at some of sources of noise you list.
These are largely unrelated to the photons impinging on your detector and can be removed by proper calibration. for example you mention crosstalk. This can be removed by using shielded twisted pair, coaxial cable and a suitable physical separation between the cables. Of course sigma algoithms can reduce the effect of cosmic ray hits
The correct calibration procedures will reduce the remaining noise to a minimum
You add shot noise to the list. now I understand that shot noise is poisson noise caused by the statistical variations inherent in the impinging photons.
This is the noise I am referring to for the sake of this debate.
Now I agree that the amount of signal received is proportional to the exposure time and the noise inherent to the impinging signal = square root of the mean photon flux over the integration time. I cannot see that the noise is going to exceed the signal over time because each summation adds the signal as well as adding the noise, and the noise is by definition substantially smaller than the signal over the integration time. (after all you did calibrate and apply the other subtraction algorithms to minimise the other sources of noise)
signal combined = signal raw minus signal dark.
Now I understand that my perspective is coming from compromise being that I use a dslr, a budget mount that is overloaded to say the least and a scope with fl=1520mm at prime focus. Hence I must consider all the compromise as a given. However, these issues still apply even to the Hubble which also uses track and stack to produce the stunning images which it does on a regular basis. Cant wait to see what the Webb does.
Regards
Steve

[Peter Ward]

Steve

I still disagree, as I suspect you are confusing signal noise ratio with noise on its own.

Total noise in an image (not just shot noise) is not by definition always smaller than the signal, and cannot always be fully calibrated out.

Very dim objects can have a signal well below the noise floor of the camera, and no amount of quadratic summing will make the signal appear.

Adding identically exposed images will increase the *S/N ratio* by the square root of the number of images. But if the S/N ratio is close to 1:1
...which is what I alluded to earlier...then you'll still end up with a noisy image.

Originally Posted by skwinty

the noise is by definition substantially smaller than the signal over the integration time.

[skwinty (Steve)]

Hi Peter
Okay I understand what you are saying.
You are talking about objects with a magnitude of say about 10 and dimmer.
Is that about right.
Now if you take darks, flats and bias frames etc and subtract these from the raw data surely that would remove most of those noise components if the temperature remains constant. You would then have a signal free of noise albeit a very small signal which would necessitate a great deal of frames to get a reasonable signal to noise ratio. Shoot so much to learn.
I was out tonight trying out my gear and am almost at wits end trying to get polar alignment. I live in a light polluted suburb, my house roof is in the way of octans and still to learn the intricacies of drift alignment.
This is not an easy hobby and not having a permanent setup compounds the issue. I think my next expenditure must be a shed with roll off roof so I dont have to go through this pain every time.
I took lots of shots but with terrible tracking and star trails plus coma. I never used my coma corrector. Am reasonably happy with the 40DH sensitivity to Ha and star colour in general. Also not having the focus extender make the fitting of the camera a precarious arrangement as I must offset the nose piece in the focuser as i was using 40mm eyepiece projection.
Theory and practice are light years apart and as much as I think I know, putting it into practice is another story.
Well I suppose perseverance will have its rewards, so I press on whenever I get an opportunity.
Thanks for your patience and input.
Regards
Steve

[anthony.tony]

Thank you both for your replies on Stacking and noise reduction Techniques. Tony