Pages: [1] 2   Go Down

Author Topic: confused myself srgb and bit depth  (Read 2707 times)

sgwrx

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 288
confused myself srgb and bit depth
« on: September 24, 2015, 02:39:59 PM »

so i can generate 16bit tif (actually i can do 32bit/channel too) files from a computer 3d modeling program.  if i save that image which is untagged, then open it in photoshop, i get the warning that there is no embedded profile.  in order to judge colors by the program's viewer window, i will assign my monitor profile and then convert to either argb or srgb.

i don't think i've ever seen this question before - but will srgb be able to contain all the computer generated colors from the renderer?  i mean, there must be a limit to actual # of colors contained within srgb as compared to say argb.  i do realize it'll be stronger in some colors and weaker in others etc.

is it all the same?  srgb = 16.7million max # of colors and argb=16.7million max # of colors? because it's all 0-255 for rgb?
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 12490
    • http://digitaldog.net/
Re: confused myself srgb and bit depth
« Reply #1 on: September 24, 2015, 02:45:55 PM »

i mean, there must be a limit to actual # of colors contained within srgb as compared to say argb.
is it all the same?  srgb = 16.7million max # of colors and argb=16.7million max # of colors? because it's all 0-255 for rgb?
The encoding or bit depth defines the number of device values (not necessarily colors). The same bit depth (8-bits per color), the same number of device values no matter the color gamut of the working space. Gamut is the range, not the number of colors or more precisely, device values. See: http://digitaldog.net/files/ColorNumbersColorGamut.pdf
Logged
Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/

sgwrx

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 288
Re: confused myself srgb and bit depth
« Reply #2 on: September 24, 2015, 03:20:51 PM »

ah ha!  perfect!  yep, i am one of those that thought a plot of a color gamut which was bigger had more number of colors!  "where did all those extra colors go if it's clamped down to 0-255" i wondered :)

the key is, understanding that 0/255/17  is a device value.  that helps me also understand the "why" of two same sets of rgb numbers can look different in two different color spaces.

excellent.  so anyway, what's great, is that i've compared the 16bit tif files generated to 8bit jpg's (both with my monitor profile converted to srgb) and the difference was huge in some areas and less in others.

i started running into this when i was working on night-renders where there were deep blue to orange/yellow sunset sky colors.  any amount of curves adjustments would make those otherwise smooth sky gradients band like a banshee, even when printing. 

thanks.
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 12490
    • http://digitaldog.net/
Re: confused myself srgb and bit depth
« Reply #3 on: September 24, 2015, 03:25:01 PM »

the key is, understanding that 0/255/17  is a device value.  that helps me also understand the "why" of two same sets of rgb numbers can look different in two different color spaces.
The scale is different. That's what a color space defines. It's as if I told you I live 1000 from you. 1000 what? Miles, feet, kilometers? The numbers without a scale are totally ambiguous.
0/255/17 in sRGB is a different device value than 0/255/17 in Adobe RGB (1998). Just as 1000 feet is a different distance than 1000 meters.
Logged
Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/

sgwrx

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 288
Re: confused myself srgb and bit depth
« Reply #4 on: September 25, 2015, 10:53:21 AM »

thanks again.

so when i use a photoshop plugin for "counting colors" or a program like irfanview to count colors in an image, it's counting device values?  so those counters look at an image and count basically all the combinations of r/g/b it finds.  even if it's different by only 1 number, it counts it.
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 12490
    • http://digitaldog.net/
Re: confused myself srgb and bit depth
« Reply #5 on: September 25, 2015, 10:59:27 AM »

so when i use a photoshop plugin for "counting colors" or a program like irfanview to count colors in an image, it's counting device values?  so those counters look at an image and count basically all the combinations of r/g/b it finds.  even if it's different by only 1 number, it counts it.
I don't know what plug-in you're referring to. It's probably counting device values. You'd need something like ColorThink Pro to extract unique colors as device values AFAIK to get the 'correct' answer. And what's the answer you're looking for? Combinations of r/g/b values or actual visible colors?
Logged
Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/

sgwrx

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 288
Re: confused myself srgb and bit depth
« Reply #6 on: September 25, 2015, 11:29:06 AM »

i'd be looking for number of unique # of r/g/b values in the image.  seems like a simple thing to do.  in my 8-bit image there are x number of unique r/g/b values, in the 16-bit image there are x...

it helps me understand why it's happening - why does the 16bit image look better, less blotchy and less pixelated in some areas upon zoom-in.  i think that's what you've explained already and in your pdf.

one image can simply have more rgb values in it than another :)  and by having this situation, it's easier to make adjustments and not run into issues like banding in gradients.
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 12490
    • http://digitaldog.net/
Re: confused myself srgb and bit depth
« Reply #7 on: September 25, 2015, 12:56:51 PM »

i'd be looking for number of unique # of r/g/b values in the image.  seems like a simple thing to do.  in my 8-bit image there are x number of unique r/g/b values, in the 16-bit image there are x...
Which are device values, not necessarily colors. See the example of the two sRGB device values that are one color in the PDF.
On-screen, in fact output to a print, a 16-bit (high bit) image shouldn't look different than it's 8-bit per color cousin unless you've edited that 8-bit per color document to the point you've introduced banding that becomes visible somewhere. And again, these are device values, not necessarily colors (color is something you have to see).
Logged
Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/

sgwrx

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 288
Re: confused myself srgb and bit depth
« Reply #8 on: September 25, 2015, 01:14:54 PM »

yep, i do get the not necessarily colors.  however to the computer are they separate colors?  id guess yes.

Sent from my XT1080 using Tapatalk

Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 12490
    • http://digitaldog.net/
Re: confused myself srgb and bit depth
« Reply #9 on: September 25, 2015, 01:22:22 PM »

yep, i do get the not necessarily colors.  however to the computer are they separate colors?  id guess yes.
If you can't see it, it's not a color. Two different device values can appear to be the same color. That's why I'm separating device value from color.
Logged
Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/

BartvanderWolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6713
Re: confused myself srgb and bit depth
« Reply #10 on: September 25, 2015, 02:43:27 PM »

If you can't see it, it's not a color. Two different device values can appear to be the same color. That's why I'm separating device value from color.

While a pair of RGB coordinates can represent 2 (individually) visible colors, they may be perceptually so close that we cannot discriminate their being different colors if placed next to each other, but they are still both colors (because they are visible separately/individually).

Some RGB coordinates can represent imaginary colors (depending on the RGB colorspace), they can not be seen by the human visual system, and should therefore be called imaginary colors, not (real) colors.

Simple color counting tools also count imaginary colors (RGB coordinates that are present in the image file data, but outside the gamut of human vision). Better tools only count Real Color coordinates, and even more elaborate tools only count those colors that can be discriminated if displayed side by side.

However, colors that cannot be discriminated when displayed side by side, may still be counted as separate colors if they happen to be displayed next to colors that are different enough to enable discrimination from each other. This point seems to be rarely mentioned.

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 12490
    • http://digitaldog.net/
Re: confused myself srgb and bit depth
« Reply #11 on: September 25, 2015, 03:23:13 PM »

While a pair of RGB coordinates can represent 2 (individually) visible colors, they may be perceptually so close that we cannot discriminate their being different colors if placed next to each other, but they are still both colors (because they are visible separately/individually).
Of course, they are both colors; the same color.  ;D
Quote
Some RGB coordinates can represent imaginary colors (depending on the RGB colorspace), they can not be seen by the human visual system, and should therefore be called imaginary colors, not (real) colors.
Right. And I prefer to call those device values, not colors as we can't see them. Or so called imaginary colors (some do have issue with that language, I'm OK with it).
Quote
Simple color counting tools also count imaginary colors (RGB coordinates that are present in the image file data, but outside the gamut of human vision). Better tools only count Real Color coordinates, and even more elaborate tools only count those colors that can be discriminated if displayed side by side.
ColorThink can do both.
Quote
However, colors that cannot be discriminated when displayed side by side, may still be counted as separate colors if they happen to be displayed next to colors that are different enough to enable discrimination from each other. This point seems to be rarely mentioned.

Again, if you can see them, they are colors.
Logged
Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/

BartvanderWolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6713
Re: confused myself srgb and bit depth
« Reply #12 on: September 25, 2015, 06:10:17 PM »

Of course, they are both colors; the same color.

If they have different a spectral composition (and different RGB colorspace coordinates), they cannot be the same color. They can be close, even too close to visually discriminate side by side, be perceived as being visually the same, but they remain different colors. Similar at best ( 0 < deltaE <= 1 ) , but not the same ( deltaE == 0 ). These invisible differences can be amplified (e.g. by profiling) and become visible, because they are not the same.

Cheers,
Bart
« Last Edit: September 25, 2015, 06:12:17 PM by BartvanderWolf »
Logged
== If you do what you did, you'll get what you got. ==

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 12490
    • http://digitaldog.net/
Re: confused myself srgb and bit depth
« Reply #13 on: September 25, 2015, 07:19:15 PM »

If they have different a spectral composition (and different RGB colorspace coordinates), they cannot be the same color.
The two device values that are the same color being discussed here are:

The Lab values are the same! In fact, ColorThink using more precision than the PS values above report the dE between the two being 0.01, they are the same color.
Quote
They can be close, even too close to visually discriminate side by side, be perceived as being visually the same, but they remain different colors. Similar at best ( 0 < deltaE <= 1 ) , but not the same ( deltaE == 0 ). These invisible differences can be amplified (e.g. by profiling) and become visible, because they are not the same.
IF they are perceived as the same to the observer, they are the same color. Unless you want to suggest color isn't a perceptual property. In any case, as far as the 'Standard Observer' is concerned, the two sRGB device values under discussion here are the same color when a dE 2000 metric is as tiny as 0.01!

Edit: I built a two pixel PS doc with the two colors specified above. In ColorThink I asked it to extract unique colors. It extracted ONE color list.
« Last Edit: September 25, 2015, 07:22:09 PM by digitaldog »
Logged
Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/

BartvanderWolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6713
Re: confused myself srgb and bit depth
« Reply #14 on: September 25, 2015, 09:15:11 PM »

The Lab values are the same!

But the (rounded to integer) Lab values are not perceptually uniform! Besides, if by changing e.g. contrast between the slightly different RGB coordinates, the rounded integer LAB values do become different, it only proofs that they were not the same, they just looked the same.

Also, sRGB [2, 255, 240] = Lab [90.3139, -53.8885, -7.7599] if rounded to 4 decimals with a Reference White of D50, and  sRGB [1, 255, 240] = Lab [90.3108, -53.9101, -7.7649] if rounded to 4 decimals with a Reference White of D50 (according to Bruce Lindbloom's CIE Color Calculator). Which shows their difference due to higher precision, and thus the silliness of the argument.

Visually the same does not mean that they are the same. They only look the same due to lack of precision of our instrument, our eyes. Here is a chart of the variable resolution of our eyes by wavelength.

Quote
In fact, ColorThink using more precision than the PS values above report the dE between the two being 0.01, they are the same color.

No, they are not the same (dE==0), but they look the same (dE<=1).

Quote
IF they are perceived as the same to the observer, they are the same color.

No, they are perceived as the same color, but they are not.

Quote
Unless you want to suggest color isn't a perceptual property.

No, that's nonsense, and I never said anything like that.

Quote
In any case, as far as the 'Standard Observer' is concerned, the two sRGB device values under discussion here are the same color when a dE 2000 metric is as tiny as 0.01!

So they are slightly different, and thus not the same (although they will look the same).

Let's not fall in the trap of, "I cannot see it so it does not exist". Closing one's eyes, or not having the resolution to see the differences will not make things stop existing.

Quote
Edit: I built a two pixel PS doc with the two colors specified above. In ColorThink I asked it to extract unique colors. It extracted ONE color list.

And the precision was rounded to, how many digits, 2? Rounding to integer values or 2 decimals may be convenient for display, but it is far from precise, and can lead to wrong conclusions.

Cheers,
Bart
« Last Edit: September 25, 2015, 09:30:33 PM by BartvanderWolf »
Logged
== If you do what you did, you'll get what you got. ==

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 12490
    • http://digitaldog.net/
Re: confused myself srgb and bit depth
« Reply #15 on: September 25, 2015, 09:27:05 PM »

The dE values are far, far, FAR less than 1! A tiny fraction of 1! They appear as the same color, therefore the are, end of story.
Logged
Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 12490
    • http://digitaldog.net/
Re: confused myself srgb and bit depth
« Reply #16 on: September 25, 2015, 11:11:40 PM »

http://www2.chromix.com/colorsmarts/smartNote.cxsa?snid=1145&-session=SessID:44231C37031a714788shs38609A6&-session=SessID:44231C37031a714779IPp385FFEB
Quote
Anyway, what delta-E (dE) IS, is a single number that represents the 'distance' between two colors. The idea is that a dE of 1.0 is the smallest color difference the human eye can see.
http://help.efi.com/fieryxf/KnowledgeBase/color/Delta%20E_H_T.pdf
Quote
0- 1 A normally invisible difference
Understanding color management by Sharma, page 101: A dE of 1 is defined as just noticeable difference. (the two device values presented here have a dE 2000 of 0.01!).
http://printwiki.org/CIELAB
Quote
Image scientists generally consider delta-E less than one to be imperceptible.

And on, and on and on...
Logged
Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/

xpatUSA

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 378
    • Blog
Re: confused myself srgb and bit depth
« Reply #17 on: September 27, 2015, 10:41:33 AM »

While the excursion into L*a*b*, Delta-E and JND is fair enough, if there is any difference at all then the colors can not be said to be "the same", IMHO.

Personally, I find it extremely difficult to consider two 8-bit colors that differ by the value of one in one of the channels as being "the same" even though those two colors are indistinguishable on my cheap monitor. Therefore, I side with folks who think that "the same" means exactly the same unless qualified otherwise.

Might it not be best to qualify the said word, e.g. "virtually the same", "effectively the same"?

Then everybody is correct  :)

Logged
best regards,

Ted

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 12490
    • http://digitaldog.net/
Re: confused myself srgb and bit depth
« Reply #18 on: September 27, 2015, 12:40:19 PM »

Personally, I find it extremely difficult to consider two 8-bit colors that differ by the value of one in one of the channels as being "the same" even though those two colors are indistinguishable on my cheap monitor. Therefore, I side with folks who think that "the same" means exactly the same unless qualified otherwise.
My take is, two different device values are of course different. In the case of the two sRGB values provided, or their conversion to lab (decimal precision as shown by CT), no question they are different values. But with a dE 2000 of 0.01, they appear as the same color! And color being a perceptual property, if the two device values appear the same, they are indeed the same color. This is why I feel it's useful to use the language where appropriate: two device values are two device values, two device values that appear the same are one color.
If we agree color is a perceptual property, if we agree that a dE value less than one, especially the two values under discussion being perceived as one color, well it's one color.
So the two device values are virtually the same but of course different, the color is effectively perceived as the same. One color.
Logged
Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/

mouse

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 253
Re: confused myself srgb and bit depth
« Reply #19 on: September 27, 2015, 03:43:02 PM »

Unfortunately this argument leads to some confusion among those of us who are not color experts. ???

If two (different) device values produce results (colors) that one can distinguish when viewed on monitor A, but are indistinguishable when viewed on monitor B: are they different colors?

If two (different) device values produce results (colors) that can be distinguished by observer X, but cannot be distinguished by observer Y: are they different colors?

Or are we reduced to the mathematical calculation of dE 2000, where if dE >= 1 they are two colors, whether or not I (or anyone else) can distinguish them on my (or any other) monitor?
Logged
Pages: [1] 2   Go Up