PixelMath Subtraction question

[melvov]

Hi,

Just wondering if there are some advance tools to help out here.
I have a very noisy DSLR image IC443 with a peculiar noise pattern. I also imaged a galaxy with a lot of background showing the same noise.
So I tried just to subtract  the galaxy image from IC443. OK sounds crazy but the pattern was gone leaving a nice silky background.
The problem is that the stars   and galaxy itself in galaxy image was also subtracted leaving black holes. The question is how to remove stars and galaxy from a linear unstretched image or how to leave just background

Mark

[georg.viehoever]

- Do your flats show the same pattern? Maybe you can use these to fix the problem.
- You may be able to create an image that does not show stars by just integrating unaligned light frames. If you did dither your light frames, you should be able to get rid of the stars, while leaving the background pattern.
- Use the search function of the forum to find some ideas for star removal. However, I think its going to be difficult to remove stars completely without introducing new artifacts.
Georg

[Carlos Milovic]

Is it a periodical pattern? Is it not corrected with darks?

[georg.viehoever]

Did you calbrate your data (flat, dark, bias)? Did you try the CanonBandingReduction script (if it is this kind of noise)? Maybe you can provide a screenshot showing the kind of patterns you have.
Georg

[melvov]

Yes, I used flats and darks.
It's not kind of noise where Canon banding script will help

Here is the image:

http://www.astrobin.com/34798/

Mark

[georg.viehoever]

Yes, I used flats and darks.

And bias? If not: Strongly recommended...

[melvov]

Well, I see but it's unlikely to add this noise.

Mark

[Carlos Milovic]

Are you talking about that medium/large scale structures that are crossing in diagonal? Because the high frequency noise seems pretty normal to me.

[melvov]

Correct, diagonal noise.

Mark

[Carlos Milovic]

Do you have flexure between the scopes, that are reflected in the same direction? Do you see these structures on the individual frames or only in the integrated one? If so, are them consistent between frames, or just random?

[melvov]

Well, it's unlikely to have flexure. It's shown in each frame.
I'm still wondering if there is a way to leave just background without stars.

Mark

[Carlos Milovic]

If it is shown in each frame, then try taking super flats. Just point near to the zenith, far away from deep sky objects, and take images as long as your light frames. Use a very high dither distance. Later, integrate the image using and aggressive rejection to high values, to delete all stars (of course, do not align them).

Another alternative could be to use Fourier analysis... but I took a look at the image, and did not see any obvious source for that noise.

BTW, I would increase the exposure times... 200" seems pretty low to me. I'm using an f/4.9 scope, and my subs are of 750" (under very dark skies)

[melvov]

Thanks Carlos,

This is an  interesting idea. I'll try.
200 seconds with LPS filter is everything I can get imaging over the city.

Mark

[Carlos Milovic]

Ok. Just make sure that your background sky level is high enough, so you are above the readout noise, and thermal noise is not that important. Also, remember that you will not be able to extract dim features without long enough exposures. It is not the same 1x 100s than 100x 1s.