Noise reduction workflow

[mmirot]

Is there a concensus on when in the work flow to preform NR techniques.  
I have reviewed the forums/examples/tutorials, there have been varing approaches.
Any tips?

Linear images or on streched images?
Both?


Other Options
Remove wavelet layer 1 prior to deconvoluton if appear to contain only noise?


Max

[Nocturnal]

Hi,

I typically do ACDNR after non-linear stretching. That makes it easier to determine the correct luminance mask.

So far the results have been good but I'd love to hear recommendations on this subject.

[Jack Harvey]

I agree Sander, a good topic for discussion. I traditionally have done my noise reductions steps in two places.  FIrst as I assemble the master frames of Red, Green, Blue, etc I use a CCDStack SIgma Reject and Hot/Cold Pixel Reject and then impute the rejected pixels.  I use ACDNR or Grey C at the the end of my image processing.  So that is my present evolution but not sure I have any solid science behind it<G>.

[twade]

Max,

My noise reduction step is the second to last step in my processing work flow.  It is just above the final sharpen step.  I wrote a tutorial on noise reduction a little over a year ago:

http://www.northwest-landscapes.com/html/articles/noise-reduction.html

This is an excellent discussion.

Wade

[mmirot]

Thank you all.
I read your excellent article Wade a while back.
The authors should upload it to the Pixinsight's tutorial page.
I am starting to like GreyRestoration too.
I wish I understood it a little more.

Max

[Juan Conejero]

Linear images or on streched images?

It depends on the image. Noise reduction is usually easier with stretched images. However, both ACDNR and ATrousWaveletTransform can work with linear images very efficiently. In this example:

http://pixinsight.com/examples/wavelets/NGC7000/en.html

I applied ATrousWaveletTransform to perform detail enhancement and noise reduction in a single process (see Step 3 in the example above).

Remove wavelet layer 1 prior to deconvoluton if appear to contain only noise?

This can only be done with oversampled images, or with low-resolution images (for example, film images, and not always).

The authors should upload it to the Pixinsight's tutorial page.

I completely agree. This tutorial is excellent. Wade, I'll write you privately to see if we can make it happen!

[Juan Conejero]

Hi Max,

I am starting to like GreyRestoration too.
I wish I understood it a little more.

GREYCstoration is a noise reduction algorithm created by David Tschumperlé, a researcher of the French CNRS. David kindly gave me his permission to implement GREYCstoration as a PixInsight module, and also assisted me in several key aspects of my initial implementation.

IMHO, GREYCstoration is simply the best general-purpose noise reduction algorithm available. It is much better than other commercial products (some of them very expensive) that are being distributed as plug-ins of a well-known imaging application. The PixInsight implementation of GREYCstoration can make use of all processor cores available, and can of course be applied to all data types (five) supported by PixInsight.

For more information, I recommend you to take a look at GREYCstoration's website:

http://cimg.sourceforge.net/greycstoration/

[Jack Harvey]

I am a big fan of Grey C