adobe RGB or s RGB for 10d

[Ron Parr]

I did the following with Phil's crayon test shots:

• Loaded both into PS
• Assigned Adobe RGB to the aRGB shot
• Converted this shot to sRGB
• Compared color values with the eye dropper

There shouldn't be much difference for colors that are in the sRGB gamut, but there are some differences that are a bit larger than what you would expect from rounding - and it's not just in the shadows. This has me thinking that either:

1. There is a bug Canon's conversion to aRGB.
2. There is a bug in Adobe's conversion to sRGB (unlikely).

3. Canon is applying a slightly different conversion process for aRGB images than it does for sRGB images. (Difference from 1 is intent.)

In the case of 3, it could be that Canon is intentionally compression the gamut to try to fit things in sRGB.

I'm assuming that you're using Canon's RAW converter or one based on the Canon SDK. It would be interesting if somebody could try reproducing this result with one of the third party RAW converters not based on Canon's SDK.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

 

[Victor Engel]

Well, I don't know what UI your RAW converter has, but after you
check off Adobe RGB as your target space, you could then go back
and check all of the other settings to make sure that they haven't
changed.

I did that. All other settings remain the same. Here is a link to my supplement:
http://www.the-light.com/Photography/dpreview/colorspace/supp.html

 

[Victor Engel]

I'm assuming that you're using Canon's RAW converter or one based
on the Canon SDK.

Yes. Breezebrowser is based upon the Canon SDK.

It would be interesting if somebody could try
reproducing this result with one of the third party RAW converters
not based on Canon's SDK.

I second this. What are the available choices? Any volunteers?

 

[Ron Parr]

I did that. All other settings remain the same. Here is a link to
my supplement:
http://www.the-light.com/Photography/dpreview/colorspace/supp.html

I noticed that before you posted it here.

This prompted me to do the check with Phil's crayons described in my other response...

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

 

[Ron Parr]

It would be interesting if somebody could try
reproducing this result with one of the third party RAW converters
not based on Canon's SDK.

I second this. What are the available choices? Any volunteers?

My understanding is that both the Photoshop converter and Phase One are based (loosely?) on David Coffin's converter, so it would be interesting to try these. Unfortunately, I don't have access to either right now.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

 

[Sheld]

I'm starting with the assumption that the RAW file source did not have clipping in it. The RAW file can be converted to either 16 bit Adobe RGB or 16 bit sRGB with no loss of information in either case. It may be possible to have slightly higher brightness without clipping when converting to Adobe RGB then when converting to sRGB, especially if the picture has some really intense greens in it.

Then the difference between sRGB and Adobe RGB is in the intensity/brightness that can be achieved in the RAW conversion without clipping. Adobe RGB supports slightly brighter pure-tone blue-greens, greens, and yellows then sRGB.

• Shel

However, in 16 bit, any of that extra gamut in the Adobe RGB image
could capture easily be preserved without any loss whatsoever by
simply turning down the intensity to avoid clipping in the sRGB
equivalent image. So that in the 16 bit format, NEITHER format
loses any information or has any inherent advantages, so the choice
of which format to use would be dictated solely by convenience
(input and output device characteristics).

As far as clipping is concerned, it doesn't matter if you're in 8
bit or 16 bit. A color that is clipped b/c it is outside of the
sRGB gamut will be clipped in either case.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

 

[Ron Parr]

I'm starting with the assumption that the RAW file source did not
have clipping in it. The RAW file can be converted to either 16
bit Adobe RGB or 16 bit sRGB with no loss of information in either
case.

No - there will be loss of information for all colors that are outside of whatever the target colorspace is - regardless of bit depth.

It may be possible to have slightly higher brightness
without clipping when converting to Adobe RGB then when converting
to sRGB, especially if the picture has some really intense greens
in it.
Then the difference between sRGB and Adobe RGB is in the
intensity/brightness that can be achieved in the RAW conversion
without clipping. Adobe RGB supports slightly brighter pure-tone
blue-greens, greens, and yellows then sRGB.

There are pretty significant differences even in the reds.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

 

[mpgoff]

Tough thread to wade through, with some advice that seems good and some that seems really bad.

I'm no pro, but my understanding is this:

• output spaces aren't exactly sRGB. If you want to get the most out of a printer, you need a printer profile, and that printer profile, even for Frontiers (one of the more famously "sRGB-like" devices), probably ain't sRGB.
• Editing in a wider space is better, right up to the point where the space is too wide. So that's no help, really, except that there are some colors in Adobe(98) that are reproduceable in some outputs, and that don't exist in sRGB.
• Color space still matters in 16-bit, which is ideal for editing anyway, especially if you are making major tonal or color moves. A great advantage of 16 bit is that it enables you to use larger spaces with less risk of posterization in the image. So, go ahead and create that masterfile in 16-bit Adobe (98), or even larger spaces, if you want (I do, but that won't influence anyone).
• For output, final corrections can/should be made according to soft-proofs with the output profile applied. If a device either really matches sRGB, or if its handlers really make it behave that way, then you can use sRGB as a "proof" and output space. In that case, if you're never going to output to anything else, or if you're truly satisfied with that output, then use sRGB all the way through.

But as usual, what the hell do I know.

• MG

 

[mpgoff]

You may want to try limited your levels or curves adjustment to Luminosity Blending mode. Then you can make independant choices about the Saturation of the image based on your creative preferences and output capabilities.

Enjoy.

• MG

I shot a picture and converted it using Adobe RGB then manipulated
it in Photoshop. One of the things I adjusted was a levels
adjustment to set white point and black point. This also enriched
the colors so they were highly saturated. I didn't need to make a
separate saturation adjustment. I found that when it came time to
print, there were large areas of the picture that were out of gamut
on the printer's profile. I printed the picture anyway, but there
were a couple of minor areas of posterization that probably
wouldn't have appeared if I had processed in sRGB to start with.

I'll have to reprocess the picture using sRGB to make sure.

Another picture where I tried both color spaces showed better
shadow detail using sRGB than Adobe RGB. This surprised me because
I expected the Adobe RGB color space to show more shadow detail.
The difference was not large, but it was noticeable.

That would be true if the color space of the CMOS would be close to
Adbove RGB. But this is not true: In fact, what the CMOS actually
records is very close to sRGB.

--
http://www.PatYuen.com

 

[santa]

When you (Sheld) say "did not have clipping in it" you use the term incorrectly. Clipping isn't a "thing" a file source has in it. It's a question of color values being beyond the color gamut or fully saturated. A RAW file has more color (wider gamut) than either sRGB or aRGB so when it is converted to either color space, some color values are lost. If you are working in 16 bit, then you certainly are going to retain a LOT more information using Adobe RGB. There is no contest. And it's not just about brightness. Adobe actually supports MORE shades. Color maps abound showing the difference. http://members.shaw.ca/johnalli/pages/tipsgamut2.html But when you talk about 8 bit people lose perspective and think that Adobe supports MORE colors and it doesn't. In 8 bit each color space supports 256 colors per channel. Adobe is just spread out more so you have to choose between a smaller gamut with colors closer together for better fine tuning of images whos colors fall in that range, and Adobe, where each color falls further from its' closest partner, but the color gamut overall is wider.

I'm starting with the assumption that the RAW file source did not
have clipping in it. The RAW file can be converted to either 16
bit Adobe RGB or 16 bit sRGB with no loss of information in either
case.

No - there will be loss of information for all colors that are
outside of whatever the target colorspace is - regardless of bit
depth.

It may be possible to have slightly higher brightness
without clipping when converting to Adobe RGB then when converting
to sRGB, especially if the picture has some really intense greens
in it.
Then the difference between sRGB and Adobe RGB is in the
intensity/brightness that can be achieved in the RAW conversion
without clipping. Adobe RGB supports slightly brighter pure-tone
blue-greens, greens, and yellows then sRGB.

There are pretty significant differences even in the reds.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

--
Canon OneDeeMarque2,TenDee, 70-2hundred 2.8IS, 16-thirty5, 1hundred
-300 5.6EL, 28-70 f3.5- something, 50m f1point4, 1.4X convrtr, tc80EN3

 

[Victor Engel]

I'm not sure I follow here. I'm inexperienced with using various blending modes. Could you describe what you mean in more detail?

The picture I was referring to was this one:

You may want to try limited your levels or curves adjustment to
Luminosity Blending mode. Then you can make independant choices
about the Saturation of the image based on your creative
preferences and output capabilities.

I got the picture just the way I wanted it, both in AdobeRGB and in sRGB. It was the printer gamut that was giving me trouble. Most of the honey jar had a gamut warning. The only part that showed posterization was right around the yellow, near the rim.

 

[jack emerson]

Yes and no. Technically yes but this doesn't mean a lot. For a
start the colours not in sRGB occur really rarely. Also, a wider
gamut with the same bit-space means the colours in that gamut are
more granular so you don't really gain anything by swapping colours
you do use for colours you don't (which is essentially what you're
doing).

Not if you use 16 bit.

Also, sRGB is extremely dense. The colors in aRGB are still pretty
densely packed, so unless you're planning on pushing the levels
pretty hard I wouldn't worry about it.

Also, if your lab or printer can't output in Adobe RGB (most are
sRGB), what exactly is the point of using Adobe RGB?

Most printers can produce colors outside of sRGB.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

I think if you read this article on differences between sRGB and AdobeRGB most of you would rethink you notions of adobe.

http://www.shootsmarter.com/infocenter/wc025.html

I always shot in AdobeRGB until I read this. Then switched. The resulting prints are vastly superior both from my printer and from photo labs.
The reasons are clearly stated in this article.
Jack

 

[mpgoff]

I'm not sure I follow here. I'm inexperienced with using various
blending modes. Could you describe what you mean in more detail?

Are you using Photoshop? Adjustment layers? If you are not using photshop, this may not help. In Pshop, you can change the blending mode of any layer in the file, including Adjustment Layers. It's a drop-down menu in the Layers Palette. Tonal moves through Levels or Curves also affect saturation, even pushing it outside your output gamut, and by changing the Blending Mode, you can reduce that effect. Specifically, changing to Luminosity blending mode should enable you to make changes to the lightness/darkness/tonality of the image without affecting the color as strongly, and as a result may spare you some of the out-of-gamut colors.

When that doesn't eliminate the gamut problems, there are other things to do (Selective Color, Hue/Sat adjustments, etc.) that can help preserve the appearance you want but still avoid bad effects in the print.

That might not have been all that clear, but hopefully it helped some. I'm not at home right now, so I can't play with the image that you've posted. Perhaps later this evening. Enjoy.

 

[cspak]

I'm not sure I follow here. I'm inexperienced with using various
blending modes. Could you describe what you mean in more detail?

Are you using Photoshop? Adjustment layers? If you are not using
photshop, this may not help. In Pshop, you can change the blending
mode of any layer in the file, including Adjustment Layers. It's a
drop-down menu in the Layers Palette. Tonal moves through Levels
or Curves also affect saturation, even pushing it outside your
output gamut, and by changing the Blending Mode, you can reduce
that effect. Specifically, changing to Luminosity blending mode
should enable you to make changes to the
lightness/darkness/tonality of the image without affecting the
color as strongly, and as a result may spare you some of the
out-of-gamut colors.

When that doesn't eliminate the gamut problems, there are other
things to do (Selective Color, Hue/Sat adjustments, etc.) that can
help preserve the appearance you want but still avoid bad effects
in the print.

That might not have been all that clear, but hopefully it helped
some. I'm not at home right now, so I can't play with the image
that you've posted. Perhaps later this evening. Enjoy.

 

[Sheld]

When you (Sheld) say "did not have clipping in it" you use the term
incorrectly. Clipping isn't a "thing" a file source has in it. It's
a question of color values being beyond the color gamut or fully
saturated.

Clipping just means blown highlights, and occurs most often in the bright whites in digital photographs. Clipping or blown highlights in a single color is less common, but also occurs, and that is where AdobeRGB helps. AdobeRGB has more bandwidth for NON-white blown highlights than sRGB. There is NO difference between AdobeRGB and sRGB for white clipping. The solution to clipping in RAW, AdobeRGB and sRGB is the same. Turn down the brightness, so that it doesn't clip. Most sensors record 12 bit values. If the value to be recorded for a given color is larger than 4095, then it clips at 4095. For 8 bit JPG files, clipping occurs at 255.

Assuming the RAW file doesn't clip, one is still left with the question of how to convert from RAW to AdobeRGB or sRGB. I stand by my original claim. If the

The conversion from RAW to AdobeRGB or sRGB using 16 bit values, than there is no information loss for either AdobeRGB or sRGB. It is possible that some RAW images can be converted do AdobeRGB of slightly higher intensity/brightness than they could be converted to sRGB. But, many images cannot benefit from this because the brighter intensity results in blown whites, and blown whites are much more common than blown pure tones. Either way, with 16 bits, all of the information is there in either AdobeRGB or sRGB. If I'm wrong, I'd appreciate it if someone explained to me what I'm not getting ... I do think I am correct in characterizing RAW clipping as just being an overexposed sensor.

• Shel

When you (Sheld) say "did not have clipping in it" you use the term
incorrectly. Clipping isn't a "thing" a file source has in it. It's
a question of color values being beyond the color gamut or fully
saturated. A RAW file has more color (wider gamut) than either
sRGB or aRGB so when it is converted to either color space, some
color values are lost. If you are working in 16 bit, then you
certainly are going to retain a LOT more information using Adobe
RGB. There is no contest. And it's not just about brightness. Adobe
actually supports MORE shades. Color maps abound showing the
difference. http://members.shaw.ca/johnalli/pages/tipsgamut2.html
But when you talk about 8 bit people lose perspective and think
that Adobe supports MORE colors and it doesn't. In 8 bit each color
space supports 256 colors per channel. Adobe is just spread out
more so you have to choose between a smaller gamut with colors
closer together for better fine tuning of images whos colors fall
in that range, and Adobe, where each color falls further from its'
closest partner, but the color gamut overall is wider.

 

[Ron Parr]

I think if you read this article on differences between sRGB and
AdobeRGB most of you would rethink you notions of adobe.

http://www.shootsmarter.com/infocenter/wc025.html

This has already been discussed in this thread.

To recap: The article doesn't change the fact that there are colors in the aRGB gamut that you can print on many devices but that cannot be represented in sRGB.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

 

[Ron Parr]

Most sensors record 12 bit values. If the
value to be recorded for a given color is larger than 4095, then it
clips at 4095. For 8 bit JPG files, clipping occurs at 255.

Sensors are analog devices. Voltages are fed to A/D converters which then produce output at varying bit depths, often 12 for most digital SLRs these days.

If the voltage going into the A/D converter is greater than the maximum amount, the value for that pixel will be clipped to 4095.

None of this has anything to do with the target colorspace yet.

Assuming the RAW file doesn't clip, one is still left with the
question of how to convert from RAW to AdobeRGB or sRGB. I stand
by my original claim. If the
The conversion from RAW to AdobeRGB or sRGB using 16 bit values,
than there is no information loss for either AdobeRGB or sRGB.

This is false. Any color that is outside of the gamuts of these spaces will necessarily be clipped regardless of bit depth.

Either way, with 16 bits,
all of the information is there in either AdobeRGB or sRGB. If I'm
wrong, I'd appreciate it if someone explained to me what I'm not
getting ...

I'm not sure where the confusion is. The gamut is the same regardless of bit depth. If color is outside of the gamut, there's just no way to express it and it must be clipped.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

 

[mpgoff]

Pardon my naivete. Help?

 

[Sheld]

I find your responses quite interesting, but we probably agree on what is really happenning.

If the voltage going into the A/D converter is greater than the
maximum amount, the value for that pixel will be clipped to 4095.

None of this has anything to do with the target colorspace yet.

....If the
The conversion from RAW to AdobeRGB or sRGB using 16 bit values,
than there is no information loss for either AdobeRGB or sRGB.

This is false. Any color that is outside of the gamuts of these
spaces will necessarily be clipped regardless of bit depth.

This is not enitirely correct. First, start by noting that sRGB and AdobeRGB define the same shades of gray and white as exactly the same triplets of Red, Green, and Blue. So for 8 bit color representations in either sRGB or AdobeRGB, the brightest white than can be expressed is 255,255,255, and that shade of white is exactly the same in both. Whites brighter than 255,255,255 are clippd. Middle gray of 128,128,128 is also the same in both color schemes. The two schemes differ in their definitions of colors where one or two of the Red/Green/Blue values is equal to 255 or close to it.

Start with a RAW image, with both whites and greens in it. The natural conversion of the RAW image might be to whites with RGB values up to 200 or so and some bright Greens with sRGB(G) values as high as 255. Converted to sRGB, the whites come out at R,G,B=200, and the greens at G=255. Converted to AdobeRGB, the white has R,G,B=200 and the green has RGB(G)=180 or so because AdobeRGB has a wider gammut. One can increase the brightness of the AdobeRGB image without clipping the Green. That way, the white is now RGB=255, and the green is also RGB(G) 250 or so. I think that is all the wider gamut of Adobe RGB means. But both formats can store the image equally faithfully, using 16 bit values. But the sRGB image could be brightened when it is converted to AdobeRGB without blowing the Green highlights, whereas if you brighten the image in sRGB format, the Green highlights would be clipped.

Now, take an 8 bit AdobeRGB image whose gamut exceeds that representable in sRGB. Perhaps it has some bright greens, whose G value is near 255, and also lets assume the image has some bright whites whose RGB value is also near 255. The white values in both images have the same RGB values. But that bright green value would want to be off the scale in sRGB, say around 320 or so. So, how about just turning down the brightness? That's equivalent to underexposing the image by a 1/3rd of a stop or so. Now the Green is no longer clipped, but the white is now more like 200,200,200 instead 255,255,255. If this whole thing is done in 16 bit land, both the sRGB image and the AdobeRGB image have ALL of the information from the original RAW image.

So, take an image in AdobeRGB

Putting that image into sRGB will lead to clipped
If the intensity of the image is turn down by 1/2 a stop (usually much less)

Either way, with 16 bits,
all of the information is there in either AdobeRGB or sRGB. If I'm
wrong, I'd appreciate it if someone explained to me what I'm not
getting ...

I'm not sure where the confusion is. The gamut is the same
regardless of bit depth. If color is outside of the gamut, there's
just no way to express it and it must be clipped.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

 

[Ron Parr]

This is false. Any color that is outside of the gamuts of these
spaces will necessarily be clipped regardless of bit depth.

This is not enitirely correct.

It really is entirely correct.

First, start by noting that sRGB
and AdobeRGB define the same shades of gray and white as exactly
the same triplets of Red, Green, and Blue. So for 8 bit color
representations in either sRGB or AdobeRGB, the brightest white
than can be expressed is 255,255,255, and that shade of white is
exactly the same in both. Whites brighter than 255,255,255 are
clippd. Middle gray of 128,128,128 is also the same in both color
schemes. The two schemes differ in their definitions of colors
where one or two of the Red/Green/Blue values is equal to 255 or
close to it.

We're in agreement here.

Start with a RAW image, with both whites and greens in it. The
natural conversion of the RAW image might be to whites with RGB
values up to 200 or so and some bright Greens with sRGB(G) values
as high as 255. Converted to sRGB, the whites come out at
R,G,B=200, and the greens at G=255. Converted to AdobeRGB, the
white has R,G,B=200 and the green has RGB(G)=180 or so because
AdobeRGB has a wider gammut.

That's not what happens at all. (200, 255, 200) sRGB gets converted to (216, 255, 201) in aRGB - with some rounding. Note that you're talking about a color that's not very saturated. You also need to take into account that the different colorspaces have different primaries.

One can increase the brightness of
the AdobeRGB image without clipping the Green. That way, the white
is now RGB=255, and the green is also RGB(G) 250 or so. I think
that is all the wider gamut of Adobe RGB means.

No - the wider gamut means that aRGB contains colors that cannot be expressed in sRGB. This is the essence of what it means for one colorspace to be wider than another.

ut both formats
can store the image equally faithfully, using 16 bit values.

Both can store 2^48 different colors, but there will be very saturated colors that aRGB can express that sRGB can't really come close to.

But
the sRGB image could be brightened when it is converted to AdobeRGB
without blowing the Green highlights, whereas if you brighten the
image in sRGB format, the Green highlights would be clipped.

The main difference for aRGB is saturation. When you view a gamut in a typical gamut plot, you are viewing something 3 dimensional (color is 3D) in a 2D plot. How do they do that? They set brightness (as defined to be the sum of the intensity of the individual channels) to be a constant, so you can look at a 2D plot where the third dimension is implicitly defined and not shown.

So, when you look at a gamut plot you are seeing a slice of uniform brightness through the 3D volume which is the colorspace.

In this view, there will be colors that are in aRGB, but outside of sRGB. By construction, all colors in the plot will have unifrom brightness, so brightness isn't an issue. The difference is saturation.

Now, take an 8 bit AdobeRGB image whose gamut exceeds that
representable in sRGB. Perhaps it has some bright greens, whose G
value is near 255, and also lets assume the image has some bright
whites whose RGB value is also near 255. The white values in both
images have the same RGB values. But that bright green value would
want to be off the scale in sRGB, say around 320 or so. So, how
about just turning down the brightness? That's equivalent to
underexposing the image by a 1/3rd of a stop or so.

Turning down the brightness won't change anything. At any given brightness level there will always be more saturated colors available in aRGB than in sRGB.

Now the Green
is no longer clipped, but the white is now more like 200,200,200
instead 255,255,255. If this whole thing is done in 16 bit land,
both the sRGB image and the AdobeRGB image have ALL of the
information from the original RAW image.

Using 16 bits also doesn't change anything other than the level of discretization. (255, 255, 0) in 8 bit is exaclty the same color as (4095, 4095, 0) in 16 bit. Anything that's clipped in 8 bit will also be clipped in 16-bit. The only thing 16 bit saves for you is very fine distinctions between the same colors that were defined in the 8 bit version of the space.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/