Author Topic: Application of Deconvolution  (Read 3954 times)

Offline Jules

  • PixInsight Guru
  • ****
  • Posts: 513
Application of Deconvolution
« on: 2012 September 27 12:32:37 »
Hi

When I process an image, say Ha, OIII, SII, or L,R,G,B after integration I apply deconvolution to each image. This can take quite a while. Is this a reasonable work flow or can I combine my three images then apply deconvolution if needed and get the same result?

or

Deconvolution on the Luminance layer, no need for RGB?

Thanks

Julian
« Last Edit: 2012 September 27 13:14:09 by Jules »

Offline Philip de Louraille

  • PixInsight Addict
  • ***
  • Posts: 289
Re: Application of Deconvolution
« Reply #1 on: 2012 September 27 14:59:01 »
Some people might say that using wavelet processing leads to better results than a decon in most cases. (Not that I want to start an angst between decon and wavelet extraction techniques.)
Personally when using LRGB, I treat the entire dataset as one layer (think of it as a Black Body spectrum)
When using narrow band, I treat each one separately because each layer, in this case, represents a different physical phenomena.
Philip de Louraille

Offline Jules

  • PixInsight Guru
  • ****
  • Posts: 513
Re: Application of Deconvolution
« Reply #2 on: 2012 September 27 15:02:42 »
Hi Phillip

We are thinking along the same wavelength with respect to narrowband (if you will excuse the pun). That is very interesting what you say about LRGB.

Regards

Julian

Offline Warhen

  • PTeam Member
  • PixInsight Old Hand
  • ****
  • Posts: 490
    • Billions and Billions
Re: Application of Deconvolution
« Reply #3 on: 2012 October 29 21:28:53 »
...When I process an image, say Ha, OIII, SII, or L,R,G,B after integration I apply deconvolution to each image... or Deconvolution on the Luminance layer, no need for RGB?

Thanks
Julian

Julian, Deconvolution is generally not done on chrominance (or separate R, G, B masters). Even if shot at full-res (1x1), they are there primarily to provide the color wash. Our eye/brain sees contrast, sharpness, and detail in the grayscale luminance and as you are shooting monochrome, I'd reserve this for L, as you've also likely invested the most time acquiring signal there. The NB channels can also be deconvolved providing they're providing real detail which will benefit by Decon's restorative magic. For me R R-L is about tightening stars and enhancing edge detail. Even when background and areas of weak signal (dim) are protected by masks I wouldn't suggest risking the enhancement of noise when it's not necessary.

I've taught processing since 2006. I always suggest we not start out by overwhelming ourselves with too much. Get L-RGB down first, then begin subtly introducing Ha in the correct way- not letting it overwhelm beautiful RGB color. Adding OIII and SII is very ambitious so take your time!
Best always, Warren

Warren A. Keller
www.ip4ap.com

Offline Jules

  • PixInsight Guru
  • ****
  • Posts: 513
Re: Application of Deconvolution
« Reply #4 on: 2012 October 30 11:41:57 »
Hi Warren

Thanks for the very interesting reply.

Excuse me for being dense but what is R R-L?

Regards

Julian

Offline Luigi

  • PixInsight Addict
  • ***
  • Posts: 140
Re: Application of Deconvolution
« Reply #5 on: 2012 October 30 19:43:45 »
I think he means 'Regularized Richardson-Lucy'
Regards,
Luigi Marchesi

Offline Jules

  • PixInsight Guru
  • ****
  • Posts: 513
Re: Application of Deconvolution
« Reply #6 on: 2012 October 31 02:22:42 »
Louis

Doh!   Thanks very much!

Regards

Julian