Topics

Somewhere over the Granger Rainbow [was: Reds print worst on my Canon ipf6400]


sj_90000@...
 

Michael said:
Granger Rainbow can be downloaded from here:
http://colorremedies.com/realworldcolor/downloads.html
Third link down - good luck with the reds !
============================================
Rex said:
They are easy to make.
. . . Change layer mode to luminosity and you have a Granger.
============================================
Hi Beat, Michael, Rex, et al.,
 
Michael: Unfortunately the Granger Rainbow that you referenced is poorly constructed. A proper one should be made from 256 tone grayscale ramps, and based on that fact, the size is also fixed. I have no idea where the 1600x1200 pixel size came from, but based on 256 tone grayscale ramps the size must be 1530x511. Also, the GR you referenced has only 196,115 unique colors, where as a proper GR has almost 400,000 unique colors. If you look at the histogram you’ll notice “combing”, where the tones are not evenly distributed. In fact if you look at the GR with the threshold tool you’ll also see the transitions from white to black are noisy. This implies that somewhere in the production process dithering was active, which compromised the tonal distribution. And looking at the histogram again you’ll also notice that there are more pixels in the shadows and highlights than the midtones. That’s wrong - the histogram shouldn’t look like a saddle. It should be flat, except for a very slight increases at the extremes due to the fact that it’s going from CieYxy to Cartesian coordinates - and the poles become elongated.
 
Rex: Your formula for creating a Granger Rainbow is incorrect. In fact there is no easy way to generate a GR with Photoshop since you need access to a *Lightness* blending mode - which PS doesn't have.
 
Luminosity(brightness) is not the same as Lightness(intensity).
 
Luminosity is based on perception, so colors are biased by how humans see color. Yellow being the brightest, then Cyan, Green, Magenta, Red, and lastly Blue, which we perceive as progressively darker.
 
On the other hand, Lightness is an absolute value that doesn't vary by perception. Usually normalized as a value from 0 to 1, but in graphics applications usually express from 0 to 255 (which is based on the fact that a byte of data can represent 256 distinct values).
 
 
I'll upload some proper targets to the files section soon. Be on the lookout for a notification.
 
If you'd like a plugin that counts unique colors; there's one here(second one up from the bottom of the page): http://www.telegraphics.com.au/sw/
 
HTH - Steve


Rex Waygood
 

Thanks Steve for the class in Grangers which I will study. I picked up that method on the net, I didn't look for other websites saying that was wrong! :-) The curse of the net.

Also thanks for the reminder of the Telematics website. I will try again with the count colours plugin, I have it working in Elements but it won't work in CC. I'll try again.

Beat.
I am going to assume your colours come from a camera.
I took 40  of my images processed "flat" in ACR with a default Adobe profile for my camera. So no sliders were moved. I took 10,000 Lab samples of each of my images processed from ACR in ProPhoto colourspace. That is 400,000 samples. I sliced and sectored the data in LCH and found the max vector in each slice/segment. That gave me an approximation to the gamut I had used in my 40 images. Adobe RGB left lots of colours/gamut out. I would guess that most photographers exceed the Adobe RGB colourspace in some areas of their own gamut. Intent is then used to get those into Adobe RGB. All my colours are represented in ProPhoto. :-)

Rex


sj_90000@...
 

Rex said:
Thanks Steve for the class in Grangers which I will study. I picked up that method on the net, I didn't look for other websites saying that was wrong! :-) The curse of the net.
 
Also thanks for the reminder of the Telematics website. I will try again with the count colours plugin, I have it working in Elements but it won't work in CC. I'll try again.
 
Hi Rex,
 
Yeah, it’s *always* best to be skeptical of information one gets from the net. In fact, years ago, when I did a deep dive on this subject, I also found many sites that said using luminosity blending was the way to go. I thought of educating them but found the number of sites daunting and didn’t pursue it. Maybe, since Dan’s list is more informed and influential than most, the word will get out. Maybe – haha! Nothing dies on the net!
 
As for the Telegraphics plugin: it is pretty old. Maybe sent them an email asking “What’s up?” Their email address is: support@...
 
Good luck - Steve


Michael Jahn
 

Hi Steve,

You wrote;

"but based on 256 tone grayscale ramps the size must be 1530x511"

Why ? Can you share a link to an example image that is downloadable ?

you also wrote

Also, the GR you referenced has only 196,115 unique colors, where as a proper GR has almost 400,000 unique colors.  

Why ? 

If RGB is 8 bits ( 255 level of grey ) for each channel - why would this not be 255 x 255 x 255 ?

Or - well - 16,581,375 unique colors ?

( yes, I know, I know, MOST of them are either WHITE or BLACK )

I have seen folks share "Granger Rainbows" like this one;

https://www.cameratrax.com/Color/color.php?BodyName=color_granger.txt 

but the one I shared has served me well when evaluating profiles.

Would love to see they one you use.

Respectfully,

Michael Jahn
2718 Cimmaron Ave
Simi Valley, CA 93065

805 416 6946


On Wed, Aug 28, 2019 at 7:07 PM 'SteveJ' sj_90000@... [COLORTHEORY] <COLORTHEORY@...> wrote:
 

Michael said:
Granger Rainbow can be downloaded from here:
Third link down - good luck with the reds !
============================================
============================================
Hi Michael,
 
Michael: Unfortunately the Granger Rainbow that you referenced is poorly constructed. A proper one should be made from 256 tone grayscale ramps, and based on that fact, the size is also fixed. I have no idea where the 1600x1200 pixel size came from, but based on 256 tone grayscale ramps the size must be 1530x511. Also, the GR you referenced has only 196,115 unique colors, where as a proper GR has almost 400,000 unique colors. If you look at the histogram you’ll notice “combing”, where the tones are not evenly distributed. In fact if you look at the GR with the threshold tool you’ll also see the transitions from white to black are noisy. This implies that somewhere in the production process dithering was active, which compromised the tonal distribution. And looking at the histogram again you’ll also notice that there are more pixels in the shadows and highlights than the midtones. That’s wrong - the histogram shouldn’t look like a saddle. It should be flat, except for a very slight increases at the extremes due to the fact that it’s going from CieYxy to Cartesian coordinates - and the poles become elongated.