Jump to content


Photo

LAB mode color noise reduction - it works!

  • Please log in to reply
14 replies to this topic

#1 jmasin

jmasin

    Mercury-Atlas

  • -----
  • Posts: 2784
  • Joined: 22 Dec 2008
  • Loc: Murphy, TX (DFW)

Posted 23 January 2013 - 10:10 PM

WOW, it works! :lol:

My apologies, I will not appropriately reference the original source, I've seen a few people mention this around the web, but I cannot recall exactly where (not all were AP references).

I tried this on my Rosette.

Here's a 1:1 crop (80% quality JPG to meet forum rules but no resize) of before and after performing a Gaussian Blur in photoshop on the (a) and (b) channels in LAB color mode. No other changes between left and right.

I guess the goal would be to not have this noise to start with, but given limited data, noise is a reality. This really did a nice job of reducing it a bit. Or, at least I think so, but I've been staring at it for a while... so open to "fresh" eyes/inputs.

Thanks!

Attached Files



#2 JWalk

JWalk

    Apollo

  • -----
  • Posts: 1388
  • Joined: 06 Jun 2010
  • Loc: San Antonio, TX

Posted 24 January 2013 - 12:02 AM

WOW!! That looks really nice.

#3 jaddbd

jaddbd

    Viking 1

  • -----
  • Posts: 877
  • Joined: 18 May 2005
  • Loc: Woodbine, MD

Posted 24 January 2013 - 12:06 AM

Subtle, but effective (as it should be).

John D

#4 alpal

alpal

    Soyuz

  • -----
  • Posts: 3695
  • Joined: 15 Jun 2009
  • Loc: Melbourne Australia.

Posted 24 January 2013 - 01:10 AM

Looks good but what is the difference between doing
a Gaussian Blur in RGB & in LAB colour?

#5 Mike Wiles

Mike Wiles

    Viking 1

  • -----
  • Posts: 950
  • Joined: 04 Feb 2009
  • Loc: Goodyear, AZ

Posted 24 January 2013 - 01:15 AM

Looks good but what is the difference between doing
a Gaussian Blur in RGB & in LAB colour?


By switching to LAB mode and doing the gaussian blur on just the A channel and the B channel, you effectively reduce the chrominance noise by blurring it. Because you're not applying the blur to the L-Luminance channel, there is no loss in sharpness of the image.

Mike

#6 Harel_Boren

Harel_Boren

    Vendor-The Boren-Simon PowerNewt Astrograph

  • -----
  • Vendors
  • Posts: 1209
  • Joined: 15 Mar 2009
  • Loc: Israel

Posted 24 January 2013 - 01:44 AM

Looks good but what is the difference between doing
a Gaussian Blur in RGB & in LAB colour?


By switching to LAB mode and doing the gaussian blur on just the A channel and the B channel, you effectively reduce the chrominance noise by blurring it. Because you're not applying the blur to the L-Luminance channel, there is no loss in sharpness of the image.

Mike


Thanks Mike,
Wouldn't the same result be achieved by applying noise reduction to the color layer in lrgb?
Cheers,
Harel

#7 alpal

alpal

    Soyuz

  • -----
  • Posts: 3695
  • Joined: 15 Jun 2009
  • Loc: Melbourne Australia.

Posted 24 January 2013 - 01:59 AM

Looks good but what is the difference between doing
a Gaussian Blur in RGB & in LAB colour?


By switching to LAB mode and doing the gaussian blur on just the A channel and the B channel, you effectively reduce the chrominance noise by blurring it. Because you're not applying the blur to the L-Luminance channel, there is no loss in sharpness of the image.

Mike


Wow - you're right -
I just tried it on an RGB image & compared it to a LAB image where I had gone to channels & applied gaussian blur to each a & b layer separately.

I won't be applying blur to RGB images anymore.

You learn something every day.

thanks
Allan

#8 Mr_T

Mr_T

    Apollo

  • -----
  • Posts: 1267
  • Joined: 06 Apr 2008
  • Loc: NJ

Posted 24 January 2013 - 06:08 AM

You can also boost your color using the a and b channel only

#9 alpal

alpal

    Soyuz

  • -----
  • Posts: 3695
  • Joined: 15 Jun 2009
  • Loc: Melbourne Australia.

Posted 24 January 2013 - 06:38 AM

You can also boost your color using the a and b channel only


I've been doing that for years but I never knew
about blurring noise with separate a & b LAB channels.

#10 vpcirc

vpcirc

    Skylab

  • ***--
  • Posts: 4100
  • Joined: 09 Dec 2009
  • Loc: Merced CA

Posted 24 January 2013 - 07:11 AM

Hallas now reccomends doing it in camera raw. When you open the image in PS, use open as and choose camera raw. you have to select the save options then reopen it. It has very fine adjustments. For a detailed video, see his photoshop series 5 tutorial. I think it's $20

#11 jmasin

jmasin

    Mercury-Atlas

  • -----
  • Posts: 2784
  • Joined: 22 Dec 2008
  • Loc: Murphy, TX (DFW)

Posted 24 January 2013 - 08:03 AM

Thanks Mike,
Wouldn't the same result be achieved by applying noise reduction to the color layer in lrgb?
Cheers,
Harel


I would think so yes.

In this case all I had was an RGB image (actually a bicolor from Ha and Oiii and a synthetic G). I had no luminance only as it were.

I could have created one from the R(G)B, but it would have inherited the noise from the R(G)B image.

I did try treating the Ha as luminance but I did not care for the results with only a bicolor.

#12 Harel_Boren

Harel_Boren

    Vendor-The Boren-Simon PowerNewt Astrograph

  • -----
  • Vendors
  • Posts: 1209
  • Joined: 15 Mar 2009
  • Loc: Israel

Posted 24 January 2013 - 01:01 PM


Thanks Mike,
Wouldn't the same result be achieved by applying noise reduction to the color layer in lrgb?
Cheers,
Harel


I would think so yes.

In this case all I had was an RGB image (actually a bicolor from Ha and Oiii and a synthetic G). I had no luminance only as it were.

I could have created one from the R(G)B, but it would have inherited the noise from the R(G)B image.

I did try treating the Ha as luminance but I did not care for the results with only a bicolor.


I understand - indeed if all you had was a RGB then this is literally a breakthrough !

Thanks for sharing!

Cheers,
Harel

#13 jgraham

jgraham

    Voyager 1

  • *****
  • Posts: 13901
  • Joined: 02 Dec 2004
  • Loc: Miami Valley Astronomical Society

Posted 24 January 2013 - 01:37 PM

For my RGB images (taken with an achromat and reseting the focus for each filter) I created a synthetic L by summing the R+G+B. The L was processed separately with the RGB layered on top of this. To smooth out the noise in the color data I applied a mild Gausian blur to the color layer. Blurring the color data is fairly common practice, I even do this with data from a color camera. This method has been around for quite a while. Still, a neat trick.

#14 Rick J

Rick J

    Fly Me to the Moon

  • *****
  • Posts: 5896
  • Joined: 01 Mar 2008
  • Loc: Mantrap Lake, MN

Posted 24 January 2013 - 04:24 PM

Lab color (note the a and b are lower case) offers a lot once you get the hang of it. I do nearly all my color work in Lab mode, only the initial combine and conversion back to RGB for the saved image take me out of it now. It has lots of power I can't duplicate in RGB mode. For instance pushing saturation can easily wipe detail or lower contrast in RGB but not if done right in Lab mode. Can be very useful in bringing out H alpha in galaxy images for instance. Learning Lab takes some effort but well worth it for me at least.

Rick

#15 alpal

alpal

    Soyuz

  • -----
  • Posts: 3695
  • Joined: 15 Jun 2009
  • Loc: Melbourne Australia.

Posted 25 January 2013 - 06:01 PM

Thanks Jon - for starting this thread - some great tips
have come out of it which I'll use to improve my processing.






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics