Noise help needed

[Star Gazer]

I'm having a great time taking pictures this year. My attempts won't win any prizes, but I'm pretty chuffed. The only dissapointment is the noise I get from my 7D.

Here's an example

http://dl.dropbox.com/u/16060721/Astro%20Help/IC434-Noise.png

Does the noise always streak across the image as shown above? I have never seen this in anyone elses images, but it is always in mine. This really troubles me!

How would you remove the noise?

The picture above was capture as follows

7D

20 exposures

180 secs

ISO 800

Stacked in PixInsight

I did not use any Darks, Flats or Bias frames. I have tried darks in the past but they never seem to make much if any difference to the streaking.

[Shibby]

Hi star gazer, did you have noise reduction turned on in the camera? I have read that poor dark frames can cause such an issue...

[Star Gazer]

Hi Shibby,

I'm not using any dark frames, also I am using RAW files, I don't think these are affected by dark frame subtraction (dfs). I think dfs is only applied to jpeg files.

It's work a check, I could be wrong

Thanks

[themos]

this is not noise, these are "artefacts", that is, a systematic bias that creeps into your final image and it's due to "hot pixels". You do need to shoot darks and process them correctly to get rid of them.

[themos]

for instance, here's a crop of a 2 hour stack from last night, without calibration frames (darks or flats). All those red splodges are the result of pixels that habitually show more signal than they should. In my case, I was not dithering so they have ended up on the same position in the registered, stacked image. In your case, I bet there was some drift of the object across the frame during the exposures so the hot pixels left a trail on your stacked image.

[Star Gazer]

this is not noise, these are "artefacts", that is, a systematic bias that creeps into your final image and it's due to "hot pixels". You do need to shoot darks and process them correctly to get rid of them.

Interesting theory! So are you saying that every red dot is a hot pixel? That means I must have a really bad sensor. Im not sure I agree with you because when I leave the hot pixels in there are no more the a few in each frame, but like the noise you can see then marching across the image. In fact you can count them and that will tell you how many images in the stack.

Regardless of what they are how can I get rid of them? You say I need darks but I have tried that and it makes no difference in DSS and PixInsight

[Star Gazer]

for instance, here's a crop of a 2 hour stack from last night, without calibration frames (darks or flats). All those red splodges are the result of pixels that habitually show more signal than they should. In my case, I was not dithering so they have ended up on the same position in the registered, stacked image. In your case, I bet there was some drift of the object across the frame during the exposures so the hot pixels left a trail on your stacked image.

If there was drift across my exposure wouldn't my stars be elongated? It's true I am not guiding but I believe my PA is very good.

What really troubles me is all of my attempts to produce stacked images look like this. I would have thought my polar alignment would be a bit variable; sometimes I just use the polar scope other times I use a drift alignment. The strange thing is the images always have this streaked noise. It's driving me nuts especially as I never see it in anyone else's image

[themos]

it could be a hot pixel or it could be a cold pixel. do yourself a favour and shoot darks. also put your frames in DSS and let it register them and then compute and show us the offsets. I bet we will see a pretty steady drift which will perfectly explain the linear artefacts in your image.

It doesn't mean you have a "bad sensor". You have a mass produced sensor with slight variability but when you stretch the image to make the faint nebulosity stand out, you also stretch these tiny differences from pixel to pixel.

if it "makes no difference when you're using darks" then you're not doing it right and we'll get to that.

You PA is probably not that good if you haven't measured it. You are shooting at quite a long focal length so it doesn't take much error to produce visible trailing.

[Shibby]

What themos is saying makes sense. The stars are slightly elongated in that direction. Over the session that drift adds up, and that's what you can see.

Until you get darks working correctly, stacking with Kappa-Sigma clipping algorithm should help get rid of the artefacts...

[themos]

Ah, yes, the Kappa-sigma should show improvement. Or the cosmetic touch up that DSS offers for your "lights"

[Jessun]

Agree with all above, but I'd throw in a large dither as well, as big as your guide program allows. (Briefly mentioned by themos already). That will suppress this structure of noise that you see. Better to 'scatter' the noise around than 'line it up' as your stack does. The overall streakiness will be greatly reduced this way and then when you hit it with your calibration frames it will look yet a great deal better.

Try it next time and compare.

[iwatkins]

I do see the same streaks in my 7D stacked images if I only use lights and over process the resulting DSS stacked images.

Defo uses darks and flats. Also do more lights.

Once into post processing just don't over stretch the image, I think this is where you might be getting them from.

Cheers

Ian

[Star Gazer]

Thanks for the advice everyone - definitely food for thought.

I have mixed success with DSS. Sometimes I get a good image other times I get something daft. For instance I managed to get an hour on the IC434 (i know that's not enough but the clouds thought it was) which I stacked in DSS. The image DSS gave was completely white. Whats annoying is I hadn't changed ay settings from the time before when I get a reasonable result.

So I decided to have a go with PixInsight. With PI I got a good image albeit with the grainy streaking. A couple of nights later I got another hour on IC434 (this time it was fog, can you believe it?), but it allowed me to stack the two sessions together to give two hours worth of subs. This time the noise was a lot less pronounced. From this I think I have learnt that most of my sessions are just too short, probably need 4 hours or more rather than just 2.

I have tried darks in the past but I have never seen them do anything significant. I've only used them in PI and DSS. PI is complicated to use, but I think my workflow was good so why the darks made no difference is anyones guess. May be I should go back to darks and try again?

I realise this is a bit of a 'how long is a piece of string' type question, but if you were shooting the Horsehead with you DSLR how many lights do you think you would take and what duration? Also you many darks?

Thanks again everyone

[themos]

The answer is always "as many as you can".

One way to help you with your registering and stacking is for you to do some work first: if you can produce 300x300 crops of your images (so that the filesizes are manageable in terms of downloading) and put them up somewhere, I can have a play with DSS. You would need to crop exactly at the same points for both your lights and darks (and any other calibration files you use: bias or flats). I know how to do this in Canon's DPP if you want help