DBE and Noise Reduction Questions

[joelshort]

I'm still familiarizing myself with PI's powerful tools and have some questions about the best practice using DBE and noise reduction when working with linear RGB images.
First, at what stage is DBE most beneficial?  Should I do DBE on the individual grayscale channels before channel combine, or after combine?  Or both?
The same basic question can also be asked of when to attack the noise.  Since different channels often have different noise characteristics, would it be better to attack the noise on individual channels before combination? 

Whatever the answers are, I'd like to know the "why" of the answer.  EX. what is the underlying reasoning for doing noise reduction in a certain way or at a certain stage?

Thanks in advance.

[cdesselles]

It seems that the general concensus is to apply DBE after channel combinations early in the linear stage.  Then follow with color calibration and (if needed) background neutralization.  Then you can attack noise prior to moving into the nonlinear phase.

[Zocky]

When I asked that question, I was answered that there is no benefit of doing DBE on separate channels. So if needed I perform DBE after combining RGB channels. As far as the noise reduction is concerned, I tend to do as little as possible, and mainly in linear stage.

[pfile]

i think the only reason to do it on each channel separately is if for some reason you have a really bad cast or stronger background signal in one channel vs. the others. DBE only has one threshold value which is applied to all 3 channels, so you may oversample the background of channels 2 and 3 because you had to raise the threshold high enough to pick up channel 1.

rob

[joelshort]

pfile-that gets at the heart of my DBE question.  Because of light pollution the background of the green channel is often somewhat stronger than the R and B channels.  And the gradient is also sometimes slightly worse as well. 

[niteman1946]

Hi Joel,
My practice is to do DBE on each R, G and B prior to combination.  In my mind I get better results.  If necessary, I'll do an additional DBE post combination.
Mark

[JoLo]

I'm with Mark, i do DBE on individual channels, i tend to see different gradients in different channels from my imaging sites.  For NR, i always do that after RGB combine, while the image is still linear.  I use MMT on linear image and recently changed my workflow and use TVGDenoise on non-linear images, if needed and with a soft touch.  Juan and Warren convinced me that TVGDenoise is the tool to use and they were right on.  Takes some practice and T/E, but it is a powerful denoising tool.

[joelshort]

I have been using MLT or MMT for linear noise reduction as well.  I haven't attempted noise reduction on non-linear images yet, but I'll take a look at TVGDenoise. 

It does make sense to me to run DBE on the individual channels, but I'm not sure if what makes sense to me is actually good practice.  Experimenting is the key!

[jerryyyyy]

Hi, have found this discussion very informative.  Been working on a heavily LPed image recently and this is the result:



I put my processing details up with the Astrobin post, but for DBE I did with DBE of each LRGB image with 20 (i.e. closely fitting) samples/row. Manually adjusting any samples with a star in them (move slightly). Did DBE twice.  I did the same set of samples on each image but used the L image for the standard as it had the most stars.