Why is it "wrong" to put Adobe RGB images on websites?

[digitaldog]

<blockquote>

<p>Do you know that ProPhoto is difficult to use, as no display or printer can show such huge gamut?<br />Do you know what happen increasing saturation on Prophoto?</p>

</blockquote>

<p>Gamut mismatch (fitting round pegs in square holes)

 

 

It IS true that the wider the granularity in a color space, the harder it is to handle subtle colors. This is why wide gamut displays that can't revert to sRGB (current LCD technology doesn't allow this.) are not ideal for all work (ideally you need two units).

 

 

There are way, way more colors that can be defined in something like ProPhoto RGB than you could possibly output, true. But we have to live with a disconnect between the simple shapes of RGB working space and the vastly more complex shapes of output color spaces to the point we're trying to fit round pegs in square holes. To do this, you need a much larger square hole. Simple matrix profiles of RGB working spaces when plotted 3 dimensionally illustrate that they reach their maximum saturation at high luminance levels. The opposite is seen with print (output) color spaces. Printers produce color by adding ink or some colorant, working space profiles are based on building more saturation by adding more light due to the differences in subtractive and additive color models. To counter this, you need a really big RGB working space like ProPhoto RGB again due to the simple size and to fit the round peg in the bigger square hole. Their shapes are simple and predictable. Then there is the issue of very dark colors of intense saturation which do occur in nature and we can capture with many devices. Many of these colors fall outside Adobe RGB (1998) and when you encode into such a space, you clip the colors to the degree that smooth gradations become solid blobs in print, again due to the dissimilar shapes and differences in how the two spaces relate to luminance.

 

 

</p>

[jacopo_brembati]

<blockquote>

<p>Simple matrix profiles of RGB working spaces when plotted 3 dimensionally illustrate that they reach their maximum saturation at high luminance levels.</p>

</blockquote>

<p>What do mean for "luminance"?<br>

Luminance is Y of CIE_XYZ or Y of CIE_xyY .<br>

For example:<br>

-start from R=200,G=200,B=200 in Prophoto<br>

-compute CIE_LCH (C=Chroma or Saturation)<br>

L=84.265563, C=0.000184, H=344.601696<br>

-compute xyY<br>

x=0.345669, y=0.358496, Y=64.577500<br>

Now increase C (C=20) without changing L and H, you get:<br>

x=0.366039, y=0.333029, Y=64.577500<br>

R=219, G=192, B=210 (rounded)<br>

As you can see luminance doesn't change if you modify Chroma, of course.</p>

<p> </p>

[digitaldog]

<blockquote>

<p>What do mean for "luminance"?</p>

</blockquote>

<p>Simple, the L*axis when viewed plotting such profiles next to LUT based, output profiles. Easy to do in ColorThink or a similar app.</p>

[jacopo_brembati]

<p>I see, L* is lightness.<br>

But again, Chroma (Saturation) doesn't change if you modify L*.<br>

Starting from Lab model, Chroma is = SquareRoot(a*a+b*b), that is the euclidean distance in (a,b) plane from L axis.</p>

[Tim_Lookingbill]

<p>jacopo,</p>

<p>Euclidean calculation can't work perfectly when the PCS is Lab which isn't a perfect color model. It has some non-uniformity in certain areas. IOW I imagine there's fudging going on in calculating a lot of this. But then the natural human visual response characteristics primarily influenced by adaptation can introduce its own non-uniformity making a level of precision you're trying to calculate for pretty much unattainable if not unuseable.</p>

<p>There was a discussion a while back here on a particular Samsung LCD that claimed it had a larger gamut than AdobeRGB after calibration upon examining its 3D model. When I got a hold of this newly made profile and assigned it to an AdobeRGB image, it didn't produce colors that suggested it was all that large of a gamut. Orangey tan brown shadows in skin tone became a maroon rust color while the pinkish peach highlights took on a sickly dull yellow instead of the usual over saturated appearance often achieved assigning AdobeRGB or ProPhotoRGB to an sRGB image. Some of this could've been caused by the Samsung software.</p>

<p>I can only conclude that all of this color matching stuff with regards to math is primarily about getting a reasonable simulation. But I still have to wonder what others are seeing on their calibrated wide gamut displays of images I've edited on my calibrated sRGB-ish display. From the comments I get on the appearance of color in my images I post here, I'm pretty close and that's all I can expect.</p>

[jacopo_brembati]

<blockquote>

<p>Euclidean calculation can't work perfectly when the PCS is Lab which isn't a perfect color model</p>

</blockquote>

<p>Tim, I know Lab is not perfectly uniform, but it is nearly uniform. I was speaking of saturation on editing software.<br>

Well there are 2 ways to change saturation:<br>

1- using HSV/HSL color model derived from RGB<br>

2- using LCH derived from Lab, derived from XYZ<br>

There is some nonlinearity in Lab, essentially at low saturation values, but method 2 is the best method to use.</p>

<p>The role of Lab as PCS has different motivations.<br>

It's not required that PCS is perceptually uniform, XYZ is as good PCS.<br>

Let me explain:<br>

A profile have to encode a color space, but we are interested in transforming images from a color space to a different one.<br>

A profile, call it ProfileA and call the associated colorspace ColorSpaceA , have to be good to go to ColorSpaceB,ColorSpaceC.........and so on.<br>

Any ColorSpaceM==>ColorSpaceN transform have to be possible.<br>

The only way to realyze that is using a Connection ColorSpace (PCS):<br>

-ProfileA contains info for ColorSpaceA=>PCS and back<br>

-ProfileB contains info for PCS=>ColorSpaceB and back<br>

In this way ProfileA and ProfileB can be used to go from ColorSpaceA to ColorSpaceB:<br>

ColorSpaceA=>PCS=>ColorSpaceB.<br>

The first transform is from ProfileA, the second transform is from ProfileB.<br>

Now, how can we chose the PCS ?<br>

We can choose PCS=XYZ, as CIE synthetized XYZ as human visual gamut. All colors are inside XYZ space.<br>

It is required that PCS is perceptually uniform?<br>

No, it is only a common space to go from a color space to a different color space. Indeed XYZ is not uniform.<br>

There is another thing that must be chosed: the white point.<br>

Any RGB color space (sRGB,Adobe 1998,ProPhoto,......) is builded fixing primaries (Red,Green,Blue), White and a compression/expansion curve (gamma).<br>

So when you go from RGB to XYZ the selected White Point enters in the conversion,for example:<br>

-for sRGB,Adobe 1998 you get XYZ_D65<br>

-for ProPhoto you get XYZ_D50</p>

<p>ICC fixed D50 as whitepoint.</p>

<p>Now the role of Lab as PCS.<br>

As XYZ have to be characterized from a white point, we know that Lab can be builded from XYZ if you fix a reference White.<br>

So Lab and XYZ are both good for PCS, and ICC make them both acceptable.</p>

<p> </p>

[jacopo_brembati]

<p>Tim,<br>

for the Samsung profile, it's difficult to say anything.<br>

But in my experience is not simple to get a "correct" profile, specially on wide gamut display.<br>

I know there are wonderful monitors (generally expensive) with calibration kit, and there are monitors that are difficult to calibrate and profile.<br>

Makers are searching to improve hardware and software solutions.<br>

We hope in the future.</p>

[digitaldog]

<p>I'm in Quito Ecudor for the week with sporadic email so when I return and have access to ColorThink I'll provide a screen capture. But you're making this more complex than it has to be.</p>

<p>Printers produce color by adding ink, working space profiles which are theoritically based on emissive devices are based on building more saturation by adding more light. These are due to the differences in subtractive and additive color models. There's the question of clipping. It's not at all hard to capture colors that are outside Adobe RGB. If you convert to this or a smaller space from a capture device that exceeds this gamut, gradations of those colors get clipped to solid blobs in these dark areas of color space. So the advantage of ProPhoto isn't only about retaining all those out-of-gamut colors it's also about maintaining the distinctions between them, so that you can map them into printable space as gradations rather than blobs. </p>

[jacopo_brembati]

<blockquote>

<p>So the advantage of ProPhoto isn't only about retaining all those out-of-gamut colors it's also about maintaining the distinctions between them, so that you can map them into printable space as gradations rather than blobs.</p>

</blockquote>

<p>It is not so simple.<br>

Gradations are saved if you are inside output gamut.<br>

When people increase saturation on ProPhoto image, the colors go quickly out of any output gamut (on ProPhoto you can reach values that are not colors too).<br>

They don't know as monitor can't show the true colors.<br>

So, gradations are lost.<br>

Worse, gamut mapping (perceptual intent) is penalized from such very saturated colors.<br>

You have to be a true expert to use ProPhoto and you need smart hardware and software solutions.</p>

[digitaldog]

<blockquote>

<p>Worse, gamut mapping (perceptual intent) is penalized from such very saturated colors.</p>

</blockquote>

<p>There is no perceptual table in the ProPhoto working space (matrix profiles don't have them). There are such options going from ProPhoto to an output space, but the that conversion has zero idea of the source space (it gets feed Lab). </p>

 

<blockquote>

<p>

 

 

<p>You have to be a true expert to use ProPhoto and you need smart hardware and software solutions.</p>

 

 

</p>

</blockquote>

Based on nearly a decade of use by myself and many other users, and the fact that ProPhoto primaries are used in every Adobe raw conversion engine, you'll forgive me if I say this is nonsense.

 

I'm off in a few minutes from Quito to the Galapagos islands to shoot for a week (raw, in ProPhoto <g>), with no net access so, I'm not ignoring the topic, I'm not able to.

[jacopo_brembati]

<blockquote>

<p>There is no perceptual table in the ProPhoto working space (matrix profiles don't have them).</p>

</blockquote>

<p>The perceptual intent is in printer profile (or may be monitor profile, if it is provided).</p>

<blockquote>

<p>Based on nearly a decade of use by myself and many other users, and the fact that ProPhoto primaries are used....</p>

</blockquote>

<p>I don't understand what are you saying.What do you think is nonsense?</p>

<p> </p>

[jacopo_brembati]

<p>if you don't believe me, you can believe ICC:</p>

 

<blockquote>

<p>ROMM RGB is the best choice for high-end printing on large gamut media, but only experts should use it</p>

</blockquote>

[scott_ferris]

<p>Dan,</p>

<p>Read this, it is an excellent <a href="http://www.cambridgeincolour.com/tutorials/color-space-conversion.htm">article on how important the rendering intent is</a> , why and what can go wrong, and explains why you are getting colour shifts. These choices can't be laid down, they are best looked at on an image by image basis depending on the colours and the extent of the gamut that is used in the file.</p>

<p>It is not about how much smaller sRGB is from Adobe RGB, obviously the colours can be replicated because after you readjust you get them where you want them, it is about how the colours are converted to their new space, the different ways this issue is worked out is what you need to work out for your images.</p>

[bill_tuthill]

Patrick, I mean that sRGB is good enough for most of us who do not produce images for CMYK printing, which is what you are saying as well. ProPhotoRGB and 16-bit RAW conversions require a lot of processing power, which my old computer does not have. For an example of 8-bit AdobeRGB not being sufficient for landscape work, see <a href="http://www.photo.net/digital-darkroom-forum/00Vl7k">this thread</a>. I may have done better with my $260 P&S!

[digitaldog]

<blockquote>

<p>ROMM RGB is the best choice for high-end printing on large gamut media, but only experts should use it</p>

</blockquote>

<p>Indeed, people who don’t have a clue about color management shouldn’t use ProPhoto RGB. Those that understand color spaces and RTFM, use products properly, no issues. I don’t know if that makes the later group experts or those that understand the proper use of the tools. </p>