Jump to content


Photo

Learning ImagesPlus

  • Please log in to reply
42 replies to this topic

#1 mmalik

mmalik

    Fly Me to the Moon

  • *****
  • Posts: 5322
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 10 June 2013 - 02:20 AM

In keeping with the tradition..., let's have a discussion in learning ImagesPlus at a deeper level. To kick off... what's qualitative and quantitative (in terms of slider movement) difference between 'Color Emphasis' and 'Decompress Color' in ImagesPlus 5.0 and 5.5, respectively? What does 0.500 default mean in comparison to 'Off' default? I notice colors are quite accentuated in 5.5 under default settings.

Related discussion here....

Note: Default settings are shown (these are sub-settings under Stretch | Digital Development menu).

Attached Files



#2 Mike Unsold

Mike Unsold

    Vendor, MLUnsold - ImagesPlus

  • *****
  • Vendors
  • Posts: 245
  • Joined: 21 May 2009

Posted 10 June 2013 - 01:01 PM

Excellent question Mike M.

First of all stretch functions used in astro imaging compress color because of their aggressive stretch curves.

Digital development is a popular initial stretch function used by many different astro imaging apps and it also compresses color when applied to a color image. K. Okano invented digital development back in the mid 1990s and noticed the color compression problem. The color emphasis controls in ImagesPlus 5.0 and earlier digital development is the exact method proposed by K. Okano to offset color compression when he first described and developed digital development.

Okano's color emphasis method does indeed introduce color into the compressed digital development stretched image. But the problem with the original digital development color emphasis is that it is not predictable and easily controlled.

Color decompression in ImagesPlus 5.5 is my solution to color compression and is very different than Okano's color emphasis. ImagesPlus 5.5 color decompression preserves color present in the image as the image is stretched and does not introduce new color that is not present in the image. Color decompression also helps to minimize star size and maintain color and contrast in bright areas of the image. ImagesPlus 5.5 color decompression is easy to control and is predictable. ImagesPlus 5.5 has two initial stretch functions, digital development and ArcSinH, with the same HDR scale and color decompression control. Digital development just uses a different stretch curve than ArcSinH. The digital development stretch curves fits some images better than the ArcSinH curve so I would try both then go with the one that gives you the best initial stretch.

The default color decompression value of 0.500 is used with both digital development and ARcSinH and was chosen since it produces a mild color decompression with most stacked but otherwise unprocessed astro images.

Mike M. lets look at an example using your Horse Head image.

First here is grayscale test image that can be used to calibrate your monitor. Make sure you can see all of the different black, gray, and white levels on your monitor otherwise the images that follow may appear to dark or light.

http://www.myastrono...ation-Image.gif

1) Initial ArcSinH stretch of the Horse Head image with HDR scale and color decompression.

http://www.mlunsold....-HH-Stretch.jpg

If you prefer less color then move the color decompression sliders towards the left. More red then move the red slider towards the right. Same ArcSinH stretch and HDR scale but color decompression turned off by moving the red, green, and blue sliders all the way to the left.

http://www.mlunsold....ColorDecomp.jpg

Note that without color decompression star are larger and pretty much have no color. Red and blue color is subdued with low saturation. Less detail in the bright blue area at the bottom of the image.


2) Curves used with soft light blend mode to enhance color and contrast slightly

http://www.mlunsold....M-HH-Curves.jpg

Horese Head after the above 2 steps

http://www.mlunsold....INEFILESAVG.jpg

At this point you could further smooth the image but you will loose real detail and replace it with smooth artifacts. I would suggest adding more exposure time to smooth the image then apply the same two steps above to get a smooth image with lots of real detail. Deconvolution can then be used in an effective way to sharpen and enhance the detail.

Mike U.

Attached Files



#3 mmalik

mmalik

    Fly Me to the Moon

  • *****
  • Posts: 5322
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 11 June 2013 - 09:18 PM

The default color decompression value of 0.500 is used with both digital development and ARcSinH and was chosen since it produces a mild color decompression with most stacked but otherwise unprocessed astro images.


Thanks for the detail response; very informative. 0.500 being the default on both, 'Digital Development' and ArcSinH, my initial thoughts are that this default setting might be bit aggressive; again just my take. I am basing this on the example you provide above and some experimentation I have done thus far. Not a big deal for experienced ones, I am just looking at it from a novice user perspective who may at first go with the defaults, that's all.

Let me ask you the same question in a different way... what would the equivalent settings (i.e., scale), regardless of the image, in ImagesPlus 5.5 in comparison to the 5.0 settings below:

Note: All are welcome to join and ask their questions, open discussion!

Attached Files



#4 Mike Unsold

Mike Unsold

    Vendor, MLUnsold - ImagesPlus

  • *****
  • Vendors
  • Posts: 245
  • Joined: 21 May 2009

Posted 12 June 2013 - 10:33 AM

Mike M.

K. Okano's digital development color emphasis implemented in ImagesPlus 5.0 is very different than color decompression in ImagesPlus 5.5 digital development and ArcSinH.

Color decompression in digital development 5.5 completely replaces color emphasis digital development in IP 5.0. No relation between color emphasis and color decompression parameters.

Mike U.

#5 mmalik

mmalik

    Fly Me to the Moon

  • *****
  • Posts: 5322
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 13 June 2013 - 01:30 AM

For comparison sake, below is a rendering of the same source data with 'Decompress Color' set to 0.000, and developed using ImagesPlus 5.5 'Digital Development'. This is to compare side-by-side the result produced above (#5913174). Thx

Attached Files



#6 mmalik

mmalik

    Fly Me to the Moon

  • *****
  • Posts: 5322
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 13 June 2013 - 01:32 AM

While we are comparing color, here is another control test done using PixInsight. Thx

Attached Files



#7 mmalik

mmalik

    Fly Me to the Moon

  • *****
  • Posts: 5322
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 13 June 2013 - 01:49 AM

And here is ImagesPlus 5.0 'Digital Development' result with 'Color Emphasis' set to what I show above (#5915960).

Attached Files



#8 Mike Unsold

Mike Unsold

    Vendor, MLUnsold - ImagesPlus

  • *****
  • Vendors
  • Posts: 245
  • Joined: 21 May 2009

Posted 13 June 2013 - 09:33 AM

Mike M.

If I apply only IP 5.5 DDP to your Horse Head I get

http://www.mlunsold....P/MM-HH-DDP.jpg

where the image on the left has color decompression turned off. The image on the right is the same DDP stretch but with color decompression set to R = 0.782, G = 0.50, and B = 0.274.

All of the last 3 examples that you posted have bloated stars with little or no color. Also H-alpha is red not orange salmon. You must be doing more than just an initial stretch.

For this image I prefer ArcSinH as shown above and at

http://www.mlunsold....P/MM-HH-DDP.jpg

since stars are smaller with color. Note the small blue star just to the left of the yellow star half way between the Horse Head and the blue nebula. Also the top right H-alpha area is red and the reflection nebula is blue with a trace of red in its lower right center. This is all done with just the ArcSinH stretch.

Mike U.

#9 Mike Unsold

Mike Unsold

    Vendor, MLUnsold - ImagesPlus

  • *****
  • Vendors
  • Posts: 245
  • Joined: 21 May 2009

Posted 13 June 2013 - 09:41 AM

Mike M. Sorry for the typo in the second image link.
The post should read as follows :


If I apply only IP 5.5 DDP to your Horse Head I get

http://www.mlunsold....P/MM-HH-DDP.jpg

where the image on the left has color decompression turned off. The image on the right is the same DDP stretch but with color decompression set to R = 0.782, G = 0.50, and B = 0.274.

All of the last 3 examples that you posted have bloated stars with little or no color. Also H-alpha is red not orange salmon. You must be doing more than just an initial stretch.

For this image I prefer ArcSinH as shown above and at

http://www.mlunsold....INEFILESAVG.jpg

since stars are smaller with color. Note the small blue star just to the left of the yellow star half way between the Horse Head and the blue nebula. Also the top right H-alpha area is red and the reflection nebula is blue with a trace of red in its lower right center. This is all done with just the ArcSinH stretch.

Mike U.

#10 mmalik

mmalik

    Fly Me to the Moon

  • *****
  • Posts: 5322
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 13 June 2013 - 06:08 PM

All of the last 3 examples that you posted have bloated stars with little or no color. Also H-alpha is red not orange salmon. You must be doing more than just an initial stretch.

This image may not be a good example of star color per se, but it seems like a decent example of background nebulosity, color saturation and widespread hue. Despite all my efforts, I am not able to come to terms with what I get with ‘Decompress Color’ defaults and/or the parameters being proposed. I wonder what other folks’ experiences are with the new color routine. Regards

#11 srosenfraz

srosenfraz

    Soyuz

  • *****
  • Posts: 3938
  • Joined: 06 Mar 2011
  • Loc: United States

Posted 14 June 2013 - 12:59 AM

FWIW, I've been using primarily the decompress color with DDP stretches (some ArcSinH stretches, too, but mostly DDP). I've been thrilled with what an excellent intitial stretch I can achieve while retaining good, natural color. With older versions of IP, I would do a DDP stretch and then later have to resaturate the image to bring color back into it. The decompress color function let's me start with a nicely saturated, well stretched image - it makes for much easier processing.

Thank you so much Mike Unsold for your ingenious solution to the color desaturation problem.

#12 austin.grant

austin.grant

    Viking 1

  • *****
  • Posts: 654
  • Joined: 18 Oct 2010
  • Loc: Shreveport, LA

Posted 14 June 2013 - 10:16 AM

Mike U. - The stars in that ArcSinH example are indeed more colorful, but it looks really unnatural. They seem to have a crescent of color, and it looks more like a weird CA than a natural color. Hard to tell if it was caused by the color preservation or the star shrinking, but I don't think it's the desirable outcome.

#13 mmalik

mmalik

    Fly Me to the Moon

  • *****
  • Posts: 5322
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 14 June 2013 - 12:56 PM

The image on the right is the same DDP stretch but with color decompression set to R = 0.782, G = 0.50, and B = 0.274.


Only way I can get some semblance of normality is if I use something like following for 'Decompress Color' [in DDP]...

R=0.202
G=0.202
B=0.098

Attached Files



#14 mmalik

mmalik

    Fly Me to the Moon

  • *****
  • Posts: 5322
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 14 June 2013 - 12:56 PM

...and the result:

Attached Files



#15 mmalik

mmalik

    Fly Me to the Moon

  • *****
  • Posts: 5322
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 14 June 2013 - 01:02 PM

If folks would like to try the same data, combined/RAW ImagesPlus CR_COMBINEFILESAVG.FIT is located here....

Note: It is the first file on the left.

#16 Mike Unsold

Mike Unsold

    Vendor, MLUnsold - ImagesPlus

  • *****
  • Vendors
  • Posts: 245
  • Joined: 21 May 2009

Posted 14 June 2013 - 01:04 PM

Hi Austin,

The crescent of color shown in the bright blue stars and also some of the brighter yellow stars are in the stacked initial image shown below. Initial blue star on left and the same star after ArcSinH stretch shown on right.

No star shrinking was done. ArcSinH just preserved the color present in the blue crescent. You can adjust the red, green, and blue compression continuously to reduce, increase, or remove the blue crescent completely. But as shown below the blue crescent is in the stacked image before ArcSinH and should also be present after the stretch.

Mike

Attached Files



#17 Mike Unsold

Mike Unsold

    Vendor, MLUnsold - ImagesPlus

  • *****
  • Vendors
  • Posts: 245
  • Joined: 21 May 2009

Posted 14 June 2013 - 01:19 PM

Mike M.

The R,G,and B color decompression sliders are intended to be adjust for each image and also personal taste. Your color decomp parameters

R=0.202, G=0.202, B=0.098

are fine but the blue value does wipe out the blue crescent on the brighter large blue stars present in your stacked image before it is stretched.

Your H-alpha is an accurate red and not the salmon orange color.

Mike U.

#18 mmalik

mmalik

    Fly Me to the Moon

  • *****
  • Posts: 5322
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 14 June 2013 - 05:49 PM

The R,G,and B color decompression sliders are intended to be adjust for each image and also personal taste.


While we all fully understand the personal taste and highly customized nature of AP processing, my initial thoughts are as follows regarding the new 'Decompress Color'; again this is just my take and is intended to provide you some honest feedback of the product, NOT a critique!

1. First and foremost, the default 0.500 is very aggressive in my opinion; at times I am 'take the defaults' kind of person and this value is pretty much a spoiler for me to say the least.

2. While preferences, as I have stated above, will vary from person to person, they should vary within reasonable limits for most AP astronomers; I am completely lost on the logic behind the two color compression settings you have tried/suggested above:

For ArcSinH:
R=1.404
G=0.500
B=1.164

For DDP:
R=0.782
G=0.500
B=0.274

Reason I could think of for this discrepancy might be that you may have been focusing too much on the star color (which is somewhat non-existent in this image) while I have been focusing on the diffuse background nebulosity and general look and feel, and contrast of the whole image.

3. Not that I was a big fan of ‘Color Emphasis' routine, I have yet to develop a comfort level (in terms of end result) for 'Decompress Color' routine. I am wondering if any thought was given to keeping 'Color Emphasis' as an option. I hope in time I'll start appreciating the new routine more than the old one.

4. A question, what color routine PixInsight uses? Reason I ask this since all renderings of the same data in PixInsight end up rustic looking... by most imagers (which I know is the not the correct color but can be pleasing to the eye at times)

5. Without turning this discussion into a comparison of PixInsight or ImagesPlus, just something I have noticed on this image is that I see the most contrast in PixInsight, followed by the most contrast in ImagesPlus 5.0 and then ImagesPlus 5.5 and it shows in my renderings in this thread. This is not a concern, rather something I have noticed that I am not getting as much out of 5.5 [using DDP that is].

6. Last but not least, would you consider setting the color decompression defaults to something that could be used out of the box; current defaults are simply too aggressive [for both DDP and ArcSinH].

Hope this analysis doesn’t come as a critique, just a learning effort. Regards


Note: I have not tried to pull color out of some crescents around stars since they seem more like chromatic aberrations from not being high in sky, and is not the star color in essence. I'll try providing some better examples of star color than this particular image.

#19 Mike Unsold

Mike Unsold

    Vendor, MLUnsold - ImagesPlus

  • *****
  • Vendors
  • Posts: 245
  • Joined: 21 May 2009

Posted 14 June 2013 - 08:04 PM

Mike M.

1) No matter what the default values are set at there will be images where the default values are too low and other images where the values are too high. This is why there are adjustable with sliders.


2) DDP and ArcSinH are very different stretch curves and the color decompress parameters are also different.

3) Okano's color emphasis is just too unpredictable and is also too heavy handed to keep.


5) Are you comparing the DDP or ArcSinH stretch function result with no other functions applied to the PI Image that results from a string of several operations?

6) No matter what the default values are set at there will be images where the default values are too low and other images where the values are too high. This is why there are adjustable with sliders.

And the Last Question :
"Note: I have not tried to pull color out of some crescents around stars since they seem more like chromatic aberrations from not being high in sky, and is not the star color in essence."

Mike M if you check your stacked but otherwise unprocessed image you will find the color crescents recorded by your camera. The whole point of color decompression is to maintain color present in the image before the stretch which both DDP and ArcSinH are doing. Check the before and after of the same star shown below.

If you check several image from NASA you will find that those stars with little blue crescents in both the unstretched and stretched version of your Horse Head really are blue.


Mike M. I have a question for you.

Why do you stack your images using average ?

Mike

Attached Files



#20 Mike Unsold

Mike Unsold

    Vendor, MLUnsold - ImagesPlus

  • *****
  • Vendors
  • Posts: 245
  • Joined: 21 May 2009

Posted 14 June 2013 - 08:15 PM

"
FWIW, I've been using primarily the decompress color with DDP stretches (some ArcSinH stretches, too, but mostly DDP). I've been thrilled with what an excellent intitial stretch I can achieve while retaining good, natural color. With older versions of IP, I would do a DDP stretch and then later have to resaturate the image to bring color back into it. The decompress color function let's me start with a nicely saturated, well stretched image - it makes for much easier processing.

Thank you so much Mike Unsold for your ingenious solution to the color desaturation problem.
"

Thank you Scott.

Mike U.

#21 mmalik

mmalik

    Fly Me to the Moon

  • *****
  • Posts: 5322
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 17 June 2013 - 12:51 AM

Why do you stack your images using average?


Simple answer would be the defaults in ImagesPlus; I normally use INCR, hence the 'Min Max Excluded Average' default of 'Automatic Processing...' routine doesn't come into play. I just end up staking ICNR RAWs with 'Combine Files...' default of 'Average'. I think there is a lesson here for all of us, program defaults DO matter. Are you implying 'Average' is NOT a good option for manually combining ICNR images?

Attached Files



#22 mmalik

mmalik

    Fly Me to the Moon

  • *****
  • Posts: 5322
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 18 June 2013 - 12:59 AM

My feel is that a very conservative decompress color is the way to go in 5.5. Note: Image in the next post was stacked from scratch using 'Min Max Excluded Average' of automatic processing.

Attached Files



#23 mmalik

mmalik

    Fly Me to the Moon

  • *****
  • Posts: 5322
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 18 June 2013 - 01:04 AM

... and the result (5.5):

Attached Files



#24 mmalik

mmalik

    Fly Me to the Moon

  • *****
  • Posts: 5322
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 19 June 2013 - 11:55 PM

For comparative analysis, below is an image that was processed using 5.0 'Color Emphasis' [all the other [processing being identical]. Note: Image in the next post was stacked from scratch using 'Min Max Excluded Average' of automatic processing.

Attached Files



#25 mmalik

mmalik

    Fly Me to the Moon

  • *****
  • Posts: 5322
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 19 June 2013 - 11:55 PM

... and the result (5.0):

Attached Files








Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics