Author Topic: Best possible result with a minimum of Data  (Read 15242 times)

astropixel

  • Guest
Re: Best possible result with a minimum of Data
« Reply #15 on: 2012 May 04 00:03:37 »
Thanks Lex. If it's not too much trouble? PM on the way.

Offline Raymond

  • Member
  • *
  • Posts: 55
    • View Profile
    • Astrophotography
Re: Best possible result with a minimum of Data
« Reply #16 on: 2012 May 04 06:48:21 »
Hi Lex
I've tried your challenge. But as DSLR-user I'm completely new to LRGB combinations. The core of M51 gets completely yellow  :o - and I can't figure out where I'm doing an error. So I'll be curious on how making the correct processing.
Raymond

Planewave 14" CDK - 10Micron GM2000HPS
William Optics 110/770 & FF | HEQ5-Pro
Omegon 66/400 | MGEN-Lacerta
Canon 500d mod

astropixel

  • Guest
Re: Best possible result with a minimum of Data
« Reply #17 on: 2012 May 04 08:18:20 »
I had the same problem with the L file. I tried LRGBCombination, ChannelCombination, Pixel Math, ImageIntegration and HDRCombination all with similar results. Eventually, resorting to extracting the L* channel from the ChannelCombination RGB and recombining with the RGB using Pixel Math.

BTW. Lex. Thanks for setting up the file share.

Rowland.

Offline dayers

  • PixInsight Addict
  • ***
  • Posts: 201
    • View Profile
    • The Orlop
Re: Best possible result with a minimum of Data
« Reply #18 on: 2012 May 04 09:09:44 »
Hi Lex,

Here's how I mangled your M51 data:

http://orlop.net/images/LRGB_mmt_hdrmt.jpg

RGB

LRGBCombination - create RGB (over the rainbow)
DynamicCrop - arbitrary crop, 2721 1556 1625 1244 11.31 degrees CW -> rgb_crop
DynamicBackgroundExtractor - subtract, 96 samples, 2 iterations. ->  rgb_crop_dbe
HistogramTransformation - adjust color balance (Shadows 0, Midtones 0.5)
BackgroundNeutralization - background .0087, .0083, .0084
SCNR - remove green
ColorCalibration - background .0052, .0048, .0052
HistogramTransformation - Shadows 10.8%, Midtones 0.429 Permanent stretch.

Luminance

HistogramTransformation - match brightness, sort of
DynamicAlignment - register RGB with Lum
LRGBCombination - dump L on RGB, saturation 25%, chrominance noise reduction on
MultiscaleMedianTransform - layers 1 / 2,  3.3, 1.0,  0.5 / 2.0, 1.0, 0.0. Noise could probably be beaten down a bit more, but this is where I quit.
HDRMultiscaleTransform - 5 layers, 1 iteration,  no deringing, no external mask

Best I am able to do. I'm sure others will be able to squeeze more out of this data. Thanks for the exercise.

Dave
Dave Ayers
  Stellarvue 80 mm refractor on CG-5 mount, Orion 50mm guide scope. Imaging camera SBIG STF-8300M, guide camera ASI120mm. PHD Guiding. Sequence Generator Pro, PixInsight.

Offline Lex

  • PixInsight Guru
  • ****
  • Posts: 555
  • We do it at night...
    • View Profile
Re: Best possible result with a minimum of Data
« Reply #19 on: 2012 May 04 13:15:14 »
Hi Lex
I've tried your challenge. But as DSLR-user I'm completely new to LRGB combinations. The core of M51 gets completely yellow  :o - and I can't figure out where I'm doing an error. So I'll be curious on how making the correct processing.
Raymond, thanks for giving it a try  :D Concerning your problem; did you add the L to the RGB in LRGB combination or inverse? Did you process L and RGB separately ( ABE, DBE, Color Calibration aso...)?
It is really a huge difference between DSLR an CCD processing I saw this when I firstly tried to get on with a DSLR one...

I had the same problem with the L file. I tried LRGBCombination, ChannelCombination, Pixel Math, ImageIntegration and HDRCombination all with similar results. Eventually, resorting to extracting the L* channel from the ChannelCombination RGB and recombining with the RGB using Pixel Math.

BTW. Lex. Thanks for setting up the file share.

Rowland.

Hi Rowland;

perhaps this could work to push the result into the right light; got to try it  >:D Cheers for the share  ;)



Hi Lex,

Here's how I mangled your M51 data:

http://orlop.net/images/LRGB_mmt_hdrmt.jpg

RGB

LRGBCombination - create RGB (over the rainbow)
DynamicCrop - arbitrary crop, 2721 1556 1625 1244 11.31 degrees CW -> rgb_crop
DynamicBackgroundExtractor - subtract, 96 samples, 2 iterations. ->  rgb_crop_dbe
HistogramTransformation - adjust color balance (Shadows 0, Midtones 0.5)
BackgroundNeutralization - background .0087, .0083, .0084
SCNR - remove green
ColorCalibration - background .0052, .0048, .0052
HistogramTransformation - Shadows 10.8%, Midtones 0.429 Permanent stretch.

Luminance

HistogramTransformation - match brightness, sort of
DynamicAlignment - register RGB with Lum
LRGBCombination - dump L on RGB, saturation 25%, chrominance noise reduction on
MultiscaleMedianTransform - layers 1 / 2,  3.3, 1.0,  0.5 / 2.0, 1.0, 0.0. Noise could probably be beaten down a bit more, but this is where I quit.
HDRMultiscaleTransform - 5 layers, 1 iteration,  no deringing, no external mask

Best I am able to do. I'm sure others will be able to squeeze more out of this data. Thanks for the exercise.

Dave

Hi Dave,

first I want to thank you for your time spent! I like your image and I think and know that your image is really better than my Rocketpropulsionprocessed rendition  >:D If there was a bit less noise it would be perfect (always bearing the low Signal in mind) perhaps a slightly heavier final Histo stretch would do the job?

Well done Dave and You are very welcome for the exercise  >:D
Clear Skies!!

Lex

______________________________________

AZEQ6 GT, TS UNC 10" f5, ASI1600mm-c

HADSO (Hagen Deep Sky Observatory)20 km W of Luxemburg City

astropixel

  • Guest
Re: Best possible result with a minimum of Data
« Reply #20 on: 2012 May 04 22:48:07 »
This is the best I could come up with for now. I left the galaxy in near the bottom right corner. View at a distance >:D ;)

Very basic.

Registration, crop, extract L* and recombine with Pixel Math or DynamicAlignment, ABE, ColourCalibration, noise reduction (clone image, invert clone, Pixel Math division, mask linear image, apply linear MultiscaleMedianTransform).

Non-linear;

HT, SCNR (green average neutral) L* mask inverted, HDRMultiscaleTransform (4 layers), L* combine with LRGBCombination (0.5, 0.1), HDRCurvesTransform (2 layers), CurvesTransform, ColourSaturation, ATWT (4 layers, 3, 2, 1, 0.5, noise reduction).

Primarily, wavelet, pixel math and saturation based - is there anything else?

Calibration frames would show this image off really well.
« Last Edit: 2012 May 05 01:10:11 by astropixel »

Offline dayers

  • PixInsight Addict
  • ***
  • Posts: 201
    • View Profile
    • The Orlop
Re: Best possible result with a minimum of Data
« Reply #21 on: 2012 May 05 12:11:50 »
Quote
This is the best I could come up with for now. I left the galaxy in near the bottom right corner. View at a distance

Pretty good result, I would say (including Junior skulking down in the corner)!  Aside from not thinking of using PixelMath for anything, our workflow isn't all that much different. With PI there are always lots of ways to skin the cat. Like I said, I didn't get very aggressive with noise-reduction.

Thanks for sharing your results and workflow.

Dave
Dave Ayers
  Stellarvue 80 mm refractor on CG-5 mount, Orion 50mm guide scope. Imaging camera SBIG STF-8300M, guide camera ASI120mm. PHD Guiding. Sequence Generator Pro, PixInsight.

Offline Lex

  • PixInsight Guru
  • ****
  • Posts: 555
  • We do it at night...
    • View Profile
Re: Best possible result with a minimum of Data
« Reply #22 on: 2012 May 05 15:12:38 »
Rowland,

Hey I think this a nice result!! What is your "formula" for the combination under PixelMath? I even still have problems while doing LRGB combination first with RGB and then adding L; sometimes it is difficult to get it work  ???

Quote
Aside from not thinking of using PixelMath for anything

@ dave:

I try not to use PM also except for narrow band data; here I think that there is no way around some Math  >:D
What time passes; I think that I will have to learn more about the options PM is offering; also in "normal" imaging...

Thanks for participating @ all !

 
Clear Skies!!

Lex

______________________________________

AZEQ6 GT, TS UNC 10" f5, ASI1600mm-c

HADSO (Hagen Deep Sky Observatory)20 km W of Luxemburg City

Offline Lex

  • PixInsight Guru
  • ****
  • Posts: 555
  • We do it at night...
    • View Profile
Re: Best possible result with a minimum of Data
« Reply #23 on: 2012 May 05 16:01:48 »
Hi gents,

I had some minutes to play; hum process  >:D

Here's what came out at the end; crazy that from one set of data there are as many possible results!!

I practically used the same processing steps than my first one; just added some HDMT; better ACDNR settings IMHO and more playarounds with CT..

Here is the result:



Clear Skies!!

Lex

______________________________________

AZEQ6 GT, TS UNC 10" f5, ASI1600mm-c

HADSO (Hagen Deep Sky Observatory)20 km W of Luxemburg City

Offline Lex

  • PixInsight Guru
  • ****
  • Posts: 555
  • We do it at night...
    • View Profile
Re: Best possible result with a minimum of Data
« Reply #24 on: 2012 May 05 16:04:01 »
BTW;

I think; really think that there is really no way around massive data acquisition.  >:D
Clear Skies!!

Lex

______________________________________

AZEQ6 GT, TS UNC 10" f5, ASI1600mm-c

HADSO (Hagen Deep Sky Observatory)20 km W of Luxemburg City

astropixel

  • Guest
Re: Best possible result with a minimum of Data
« Reply #25 on: 2012 May 05 17:16:04 »
Quote
Pretty good result, I would say (including Junior skulking down in the corner)!  Aside from not thinking of using PixelMath for anything, our workflow isn't all that much different. With PI there are always lots of ways to skin the cat. Like I said, I didn't get very aggressive with noise-reduction.

This is the right approach of course - http://pixinsight.com/forum/index.php?topic=4068.msg29346#msg29346.

I mean to try this on Lex's data. Just to learn the tools.

I had another look at my work flow. Keep in mind that I am forgetful :o So, I tried it again.

Once the the RGB image is cropped - ABE and Colour Calibration (preview1 the entire galaxy as the white reference and preview2 the background).

Extract L* and recombine using PM addition, or use DynamicAlignment - pixel level only (Harry's videos). For some reason the supplied L channel is problematic? I must have a closer look at that..

Noise reduction.

Duplicate the RGB image and invert the clone.

Use the L* as a luminance mask on the linear RGB image.

PixelMath RGB/inverted clone - This is just a hack, but it reduces the background noise. Then apply MultiscaleMedianTransform for effective noise reduction. You could use ATWT in linear mode.

Don't be too aggressive with the post processing - gently, gently. It really needs calibration frames. Dithering would produce a less noisy image and improve the sub-pixel sampling too.
« Last Edit: 2012 May 05 20:51:08 by astropixel »

Offline Alejandro Tombolini

  • PTeam Member
  • PixInsight Jedi
  • *****
  • Posts: 1260
    • View Profile
    • Próxima Sur
Re: Best possible result with a minimum of Data
« Reply #26 on: 2012 May 09 22:23:14 »
Hi Lex, this is the result with your data.

I would appreciate if you can check some parts of the processing where I have lots of doubts. (I know are basic doubts but I am new combining  CCD images). For example; Luminance seems not to be a linear image, I have stretched RGB and combine whit Luminance without using HT on it.
It also surprised me the kind of noise in R, G, and B. it is like a pattern, uniform to big scale but not in deep. I am sure it is normal but very different to what it used to see in Canon.

StarAlignment with L as reference


DBE to L (Notice that I am working with out STF)


HE to a Clone


CBR to see results in the clone


Delete the clone and CBR to image


Open B channel and DBE


B channel after DBE and the noise I was talking about


Open G channel and DBE


G channel after DBE


Open R channel and DBE


R channel after DBE


RGB Combination


Whit AberrationInspector generate a the preview for BN



CC


RangeMask


Increasing protection of RangeMask with MT


Convolution before and after protecting stars with RangeMask



STR>HT


LRGBCombination


SCNR to Green


Noise reduction. MMT protecting the galaxy with RangeMask


HDRMT protecting galaxy core with RangeMask


New HDRMT protecting with new Range Mask


HT


LMask applying HT and HTWT on inversed L


CS protecting background with LMask


CT protecting background with LMask


A little noise reduction in Chrominance with ACDNR


MMT with RangeMask


HT protecting background with RangeMask


MorphologicalSelection with starmask


Still too noisy


HT to inverted L


Noise reduction. Convolution


HT


CT


Resample


Final Image

Offline Raymond

  • Member
  • *
  • Posts: 55
    • View Profile
    • Astrophotography
Re: Best possible result with a minimum of Data
« Reply #27 on: 2012 May 09 23:48:35 »
Hi Alejandro
It's the second time I'm using your workflow to improve my very limited skill of processing images.
Thanks a lot for providing a detailed description!
Raymond

Planewave 14" CDK - 10Micron GM2000HPS
William Optics 110/770 & FF | HEQ5-Pro
Omegon 66/400 | MGEN-Lacerta
Canon 500d mod

astropixel

  • Guest
Re: Best possible result with a minimum of Data
« Reply #28 on: 2012 May 10 01:54:29 »
Me too. It's great to have this level of demonstration.

Offline cs_pixinsight

  • PixInsight Addict
  • ***
  • Posts: 156
    • View Profile
Re: Best possible result with a minimum of Data
« Reply #29 on: 2012 May 10 13:00:51 »
Alejandro, you are turning out some awesome workflow tutorials.  They have been very helpful and I wanted to let you know it's greatly appreciated.

Craig