Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

MUREDenoise--any downside?

  • Please log in to reply
5 replies to this topic

#1 diggy

diggy

    Messenger

  • -----
  • topic starter
  • Posts: 443
  • Joined: 09 May 2004
  • Loc: Vancouver Island, Canada

Posted 14 August 2019 - 01:59 PM

I've been amazed at how good MUREDenoise is for linear images...it's my first step after integration.  It took me a while to get the camera parameters correct, but once dialled in, this process is like magic!  Not only is it highly effective in maintaining faint structure, but it's very easy to use (compared to TGVDenoise, etc.) 

 

But being the sceptic that I am (there's no free lunch!), I wonder if there's some down-stream downside that I'm missing.  I think I occasionally see a trace of a kind of "checker board" pattern in later non-linear stages, but I can't be sure if that's just insufficient integration time, or some other artifact.

 

What are other people experiencing?



#2 lucam

lucam

    Viking 1

  • -----
  • Posts: 635
  • Joined: 09 Oct 2017
  • Loc: Upstate NY, USA

Posted 14 August 2019 - 02:13 PM

Yeah, i think you have it right. When I go a bit overboard with MUREdenoise, I end up with what look like JPEG stair-step artifacts in later stages. When you blink it, you can see that the stairsteps where there in the original image but buried in Poisson noise. I find that backing off slightly and leaving a tiny bit of noise (often using values below the suggested variance scale) and smoothing some of the other sources of noise (pattern or otherwise) with light use of MLT gives me very clean data.


  • RossW likes this

#3 cfosterstars

cfosterstars

    Mercury-Atlas

  • *****
  • Posts: 2659
  • Joined: 05 Sep 2014
  • Loc: Austin, Texas

Posted 14 August 2019 - 06:47 PM

Just like all other noise reduction processes and almost any other process in PI, you can go overboard. I always try and err on under correcting as much a possible. I try to leave some noise in the image or it looks too plastic. But this is not a downside to MUREDenoise since it is not something that is wrong with the tool, but with how you use it.


Edited by cfosterstars, 14 August 2019 - 07:55 PM.

  • lucam likes this

#4 kingjamez

kingjamez

    Vanguard

  • *****
  • Posts: 2160
  • Joined: 03 Oct 2006
  • Loc: Fairfax, VA

Posted 14 August 2019 - 06:59 PM

Just like all other noise reduction processes and almost any other process in PI, you can go overboard. I always try and err on under correcting as much a possible. I try to leave some noise in the image or it looks too plastic. But this is not a downside to MUREDenoise since it is not something that is wrong with the tool, but with how use use it.


Exactly this. My mantra in making decisions on how far to “push” is to go as far as I think looks good, then back off until I just begin to think it needs more.

-Jim

#5 astrovienna

astrovienna

    Mercury-Atlas

  • *****
  • Posts: 2811
  • Joined: 04 Dec 2006
  • Loc: The NoVa White Zone

Posted 14 August 2019 - 10:11 PM

I love Mure Denoise, but I also find sometimes that I need to back it off a bit. Once in a while the default setting puts something like a very fine moire pattern in the background.  Easing back just slightly fixes this, and leaves a nice noise free image.

 

Kevin


  • kingjamez likes this

#6 cfosterstars

cfosterstars

    Mercury-Atlas

  • *****
  • Posts: 2659
  • Joined: 05 Sep 2014
  • Loc: Austin, Texas

Posted 14 August 2019 - 10:56 PM

I love Mure Denoise, but I also find sometimes that I need to back it off a bit. Once in a while the default setting puts something like a very fine moire pattern in the background.  Easing back just slightly fixes this, and leaves a nice noise free image.

 

Kevin

The value for the Variance scale that is calculated for the script is just a starting value. I almost never use it. I test ,many values for the variance scale to find what looks best and sometimes the best value is much higher or much lower than the starting point. I have had cases that the starting value was about 1, but I needed about 8. Other times, the starting value was one and I had to back off to 0.3. There is nothing magical about the starting estimate and it is just that - an estimate. I create about 6 previews of the same region so that I can apply a bunch of different values for the variance scale to test and then close the script to inspect them. This helps me decide what I need to use.


  • RossW likes this


CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics