Author Topic: Noise reduction comparisons  (Read 1447 times)

Offline ribuck

  • Newcomer
  • Posts: 29
    • View Profile
Noise reduction comparisons
« on: 2017 March 03 04:52:44 »
Hi All,

I'ev read thought various tutotrials and TGVDensoise seems to be commonly reffered to as the champion of noise reduction.  I've played around with the TGV quite a bit, but i always seem to be able to get better results with MultiScaleLinearTransform, followed by a tickle of ACDNR.

Are there any benefits of using TGVDenoise over MultiscaleLinearTransform or is it simply a case of using whathever works best for the individual.

Rich.
« Last Edit: 2017 March 03 05:04:53 by ribuck »

Offline eganz

  • Member
  • *
  • Posts: 50
    • View Profile
    • Eric Ganz Flickr
Re: Noise reduction comparisons
« Reply #1 on: 2017 March 03 07:40:55 »
Rich,

There are many noise reduction modules available in PixInsight, which are applicable in different situations.

For more specific response, you might want to post examples of the images and noise that you are trying to work with.

In general, one might use MLT on linear images to provide a quick and easy result. Mure is an alternative here.

I would suggest that you use TGV on stretched and composite images near the end of processing. In this case, TGV will sometimes produce remarkable results on your images.
For example, you can Separate the luminance and chrominance operations which is often very desirable.

However, TGV can be more difficult to use, and may require more experimentation on previews to get it to work as you learn how to use it. It is worth the trouble if you want the best results possible.

if you have an image where TGV is not working for you, you can try MMT on a stretched image.

In all cases, you need to pay attention to the masks and parameters to apply the noise reduction to the places you want in the right amounts.

Eric

Offline ribuck

  • Newcomer
  • Posts: 29
    • View Profile
Re: Noise reduction comparisons
« Reply #2 on: 2017 March 03 16:55:29 »
Hi eric,

the file i'm messing with is just a Lum sub of M35, which has been calibrated, Drizzle integrated,  so just stars in the image, nothing exciting.  My image is in the linear stage and i've cloned the image and applied it as a non-linear inverted mask before applying TGVDenoise and the results are quite horrific, so not sure where i'm going wrong.

I've tried applying TGVDenoise to the image in both Linear & Non-Linear state and get same results.

Rich.

 


Offline msmythers

  • PTeam Member
  • PixInsight Jedi
  • *****
  • Posts: 1176
    • View Profile
    • astrobin
Re: Noise reduction comparisons
« Reply #3 on: 2017 March 03 19:03:34 »
Rich

Try dropping your Strength to 1 and your iterations to 3 or 4.

Here is an example of my own M35 image using a non-linear inverted mask. You can see my Statistics for the image used to set the edge protection.

I normally use a non-released version of the TGV module using the 50 percent gray mask technique. That version of the module can be found here but is only available for Windows and Linux computers.
http://pixinsight.com/forum/index.php?topic=8970.0

You might like the Mure Denoise script. I think some are having success even with Drizzle integrated files. You can read more about Mure here. http://pixinsight.com/forum/index.php?topic=9206.0


Mike

Offline ribuck

  • Newcomer
  • Posts: 29
    • View Profile
Re: Noise reduction comparisons
« Reply #4 on: 2017 March 04 00:05:01 »
Hi Mike,

many thanks you were spot on. It turns out the problem was down to the number of iterations. I reduced it to down to 7 and then played with the settings a bit and it's now working perfectly with just right amount of noise reduction.

Once again, many thanks.

Rich.

Offline msmythers

  • PTeam Member
  • PixInsight Jedi
  • *****
  • Posts: 1176
    • View Profile
    • astrobin
Re: Noise reduction comparisons
« Reply #5 on: 2017 March 04 16:45:19 »
Rich

Glad that helped.


Mike