M74 How to get rid of the noise

[Ivor]

Not one to phased by a challenge my last target was M74 in Pisces, with a magnitude 10 I knew I'd be pushing my FLT110 and unmodified EOS400D but that's part of the fun.

Thanks to Carole's advice the other night I've managed to get a lot more details out of my image (I know I've pushed the blue too much) however there is a lot of noise in the background which I can't get rid of any advice gratefully received.

http://gallery.orpington-astronomy.org.uk/displayimage-1238.html

[Carole]

Hi Ivor, if you right click on properties in the gallery you can paste the URL in between the (Img) (img) when you click on the left icon above.

I've done it for you.



Noise is a constant problem with DSLRs but with your technical know how and help on here I am sure you will find a solution.  I think getting a balance between noise and how much to stretch an image is one of the quandaries.  

Carole

[MarkS]

Ivor,

It would be good to know some details of your acquisition:
How many exposures did you take?
What duration were they?
Which ISO setting did you use?

The 2 main sources of noise are
1) Noise from the background flux (light pollution)
2) Thermal noise from the camera (dark current) related to the CCD temperature

Another source is read noise from the CCD but this can be drowned out with longer exposures.

There are certainly ways to reduce noise in the final image - for instance Photoshop has a noise reduction filter and IRIS has an Adaptive Filter under the Processing menu.  But these only help to a limited extent.  The danger is that image detail gets lost during the "smoothing" process.

One final tip for the 400D is the following - make sure the rear LCD display is switched off as you take your exposures.  Otherwise the LCD heats up the CCD and makes a significant contribution to noise.

Mark

[Ivor]

I put the details on the gallery, I'll know for next time to include then here as well. I stacked in DSS with 19 x 5m Lights, Darks 15 x 5m at ISO800 & 28 flat & 30 Offset (I've decided to give up on the dark flats).

I didn't know it was possible to turn off the display completely I've been just reducing the display to as dim as possible. I shall look into that.

I've tried to use the noise filters in CS4 but it did make any difference.

I guess one of big questions how to decided the quantity of subs, for something this faint should i be using more?

[MarkS]

The faint arms of galaxies are notoriously difficult to bring above the noise level, unless you live live in a properly dark place.  I would strongly recommend that you start off by imaging bright objects such as the Pleiades, Orion Nebula, Dumbbell Nebula, various star clusters etc.  This way you can gain confidence by producing stunning images of well known objects rather than becoming disillusioned with mediocre results from things difficult to image.

If you're interested in analysing the source of the noise in your images, then it can be done by knowing that:
Bias frames contain read noise
Dark frames contain read noise + thermal noise (assuming you took them at the end of your imaging session)
Light frames contain read noise + thermal noise + background sky flux noise

So,
Take 2 successive bias frames and subtract one from the other (if your application does not allow negative pixel values then add an offset to the one of the frames to prevent the numbers being clipped)
Calculate the standard deviation of pixels in a small area of the resulting frame.  This is the read noise.

Take 2 successive dark frames and do the same thing.  This gives the std dev of the read + thermal noise.
The thermal noise is then square_root( (read+thermalnoise)^2 - (read noise)^2)  where ^2 means "square this number".

Take 2 successive light frames and do the same thing.  But make sure you choose a background area of sky with no stars for your std dev calculation.  This gives the std dev of read+thermal+background noise

The background noise is then square_root( (read+thermal+background noise)^2 - (read+thermal noise)^2 )

If you're a bit unsure of the final calculations then just post the std dev of the pixels in the 3 difference frames (bias, darks, lights) and we can do the calcs for you.

Understanding the size of the various components of noise is the first step to being able to tackle them.

Mark

[Ivor]

Thanks for the guidance, I'm not sure DSS will let me subtract the frames like that, will IRIS ? If so I'll download it and have a play.

Regarding thermal noise until now I've been leaving a gap of 30secs between each sub is this too short? If so what is the recommend gap?

[RobertM]

Carole raised the question about thermal noise and a delay last week and I think you'll find that the consensus was to ignore it as it's more beneficial to use the time to grab more subs.

Robert

[MarkS]

IRIS will definitely do that for you.

I don't think there is anything to be gained by leaving an interval between subs.

[The Thing]

Thanks for the guidance, I'm not sure DSS will let me subtract the frames like that, will IRIS ? If so I'll download it and have a play.

Regarding thermal noise until now I've been leaving a gap of 30secs between each sub is this too short? If so what is the recommend gap?

CS4 - load the images as layers and use the Subtract option to blend the layers, then flatten and do the measurements (don't know if you can do that bit in CS4).

[MarkS]

Duncan,

Can you do it in Photo Filtre?

[The Thing]

Unfortunately the free version of Photo Filtre doesn't support layers or image combination using a function. However it is free and portable (no install required).