Perceptual v Colorimetric v scene referred v output referred

[Doug Gray]

Perceptual and output referred both transform non-linearly. Colorimetric and scene referred transform linearly.

There are two main color processing paths in photography. Perceptual and Colorimetric. 99% is Perceptual but it also distorts colors and tone in order to create pleasing images and also map large dynamic ranges in nature into either displayable or printable ranges. Colorimetric retains the exact color and tone, with the  exception of white point transforms for differing illuminants. This is appropriate for reproduction purposes like taking a picture of a painting where you want to print a replica. But is completely wrong for normal photographs. And its not just dynamic range. Even when the output display or media are within the dynamic range, colorimetric processes typically produce unattractive images.

Most of the non-linear mapping of colors to produce pleasing prints is done by in camera or RAW processing tools and generates what is referred to as "output referred" images. A further, but much smaller, non-linear process transforms from the output referred image to output media. The latter accommodates a small (usually) dynamic range reduction in prints by scaling to the black point and increasing contrast in the midrange to produce a print that looks reasonably close to the output referred image on a monitor. This is what choosing "Perceptual Intent" during printing accomplishes. Alternately, one can accomplish the same but with more control by soft proofing using curves in Photoshop and printing using Colorimetric Intent.

For reproduction work you want to, if possible, precisely reproduce tone and colors. For instance if you want to make a duplicate of a comic book cover for marketing reasons you want to use colorimetric scene referred processing and either Relative of Absolute intent when printing. Absolute will retain the exact color of the reproduced art like the yellowing of an old cover. Relative intent will transform the white point which can be desirable in catalogs of new comics. For duplication, Relative intent can create somewhat darker images than the original since the white point is mapped to the print media so, for instance, a color with an L value of 80 might print with an L value of 76.

[BradFunkhouser]

...For duplication, Relative intent can create somewhat darker images than the original since the white point is mapped to the print media so, for instance, a color with an L value of 80 might print with an L value of 76.

Just to expound on your point a bit, based on my experiences...

Since the whites in a scene referred image are relative to a perfect white diffuser, they're often no whiter than L95 or L96 in relation to the L100 in a working space.  That image is in a state to be printed using absolute colorimetric.  If you print the image using relative colorimetric to a paper with a white point below L100 (which is all papers), then the whites in the image will get pushed down to the extent that the paper white is below L100 (as you describe above), and other light colors will get pushed down relatively too. 

So, from my experiences, for a scene referred image (one that's in an absolute colorimetric state), before printing relative colorimetric, you need to scale the image's whites up in the working space so they don't get pushed down too far by the relative mapping to the white point of the paper.  Likewise, if black point compensation is going to be used, the image's blacks need to be scaled down in the working space in preparation for the compensation from working space L0 to the paper's black point.

And you have to make choices and perform these scaling changes prior to exchanging the image with anyone else since they'll most likely be using relative colorimetric or perceptual rendering intents, unless they specifically know how to deal with an absolute colorimetric image and know that that's what you're sending them.

Does this match your experiences?

[Jim Kasson]

Perceptual and output referred both transform non-linearly. Colorimetric and scene referred transform linearly.

There are two main color processing paths in photography. Perceptual and Colorimetric. 99% is Perceptual but it also distorts colors and tone in order to create pleasing images and also map large dynamic ranges in nature into either displayable or printable ranges. Colorimetric retains the exact color and tone, with the  exception of white point transforms for differing illuminants. This is appropriate for reproduction purposes like taking a picture of a painting where you want to print a replica. But is completely wrong for normal photographs. And its not just dynamic range. Even when the output display or media are within the dynamic range, colorimetric processes typically produce unattractive images.

Most of the non-linear mapping of colors to produce pleasing prints is done by in camera or RAW processing tools and generates what is referred to as "output referred" images. A further, but much smaller, non-linear process transforms from the output referred image to output media. The latter accommodates a small (usually) dynamic range reduction in prints by scaling to the black point and increasing contrast in the midrange to produce a print that looks reasonably close to the output referred image on a monitor. This is what choosing "Perceptual Intent" during printing accomplishes. Alternately, one can accomplish the same but with more control by soft proofing using curves in Photoshop and printing using Colorimetric Intent.

For reproduction work you want to, if possible, precisely reproduce tone and colors. For instance if you want to make a duplicate of a comic book cover for marketing reasons you want to use colorimetric scene referred processing and either Relative of Absolute intent when printing. Absolute will retain the exact color of the reproduced art like the yellowing of an old cover. Relative intent will transform the white point which can be desirable in catalogs of new comics. For duplication, Relative intent can create somewhat darker images than the original since the white point is mapped to the print media so, for instance, a color with an L value of 80 might print with an L value of 76.

I find some or your terminology confusing. In an attempt to take the issue apart a bit, I offer the following:

Conversion of raw image values to an input-referred XYZ form (or any other linear RGB color space) using a compromise matrix is linear using most adaptation algorithms.
Conversion of raw image values to an input-referred XYZ form using a 3DLUT is in general nonlinear.
Conversion of raw image values to an input-referred sRGB form using either a compromise matrix or a 3DLUT is nonlinear.
Conversion of input-referred XYZ values to output-referred XYZ values is in general nonlinear.
Conversion of output-referred XYZ values to output-referred XYZ values in the presence of gamut limitations, using Absolute Colorimetric, is linear for in gamut colors, and nonlinear for out of gamut colors.
Conversion of output-referred XYZ values to output-referred XYZ values in the presence of gamut limitations, using Relative Colorimetric, is possibly linear for in gamut colors, and nonlinear for out of gamut colors.
Conversion of output-referred XYZ values to output-referred XYZ values in the presence of gamut limitations, using Perceptual, is likely nonlinear for in gamut colors, and nonlinear for out of gamut colors.
Conversion of output-referred XYZ values in the presence of gamut limitations to colorants of a particular printer is in general nonlinear.

I've used XYZ as a stand-in for any linear RGB color space in the above, and sRGB as a stand-in for any nonlinear RGB color space. Does that make sense to you?

Jim

[Doug Gray]

Just to expound on your point a bit, based on my experiences...

Since the whites in a scene referred image are relative to a perfect white diffuser, they're often no whiter than L95 or L96 in relation to the L100 in a working space.  That image is in a state to be printed using absolute colorimetric.  If you print the image using relative colorimetric to a paper with a white point below L100 (which is all papers), then the whites in the image will get pushed down to the extent that the paper white is below L100 (as you describe above), and other light colors will get pushed down relatively too.

When doing repro work I use a linear transform in Photoshop Raw and adjust the luminance, temp, and tint to produce the same color at whatever is close to a whitest point in the artwork. I require the actual white point on media to exceed that of the scene. This is almost always the case. And this provides the most accurate repro and is essentially automatic with Abs. intent. Rel. Intent I use less frequently but is applicable for catalogs as in my example. For that, I generally just scale L upward after the fact.

So, from my experiences, for a scene referred image (one that's in an absolute colorimetric state), before printing relative colorimetric, you need to scale the image's whites up in the working space so they don't get pushed down too far by the relative mapping to the white point of the paper.  Likewise, if black point compensation is going to be used, the image's blacks need to be scaled down in the working space in preparation for the compensation from working space L0 to the paper's black point.

I don't scale BP as the scene referred image doesn't typically dip below what the media can reproduce. If not, I will tweak curves in PS as little as possible to remove blocking. It's not colorimetric but it can produce satisfactory results.

And you have to make choices and perform these scaling changes prior to exchanging the image with anyone else since they'll most likely be using relative colorimetric or perceptual rendering intents, unless they specifically know how to deal with an absolute colorimetric image and know that that's what you're sending them.

Does this match your experiences?

If they don't understand how to use Absolute Intent they are not of interest to me. For interchange, I find it easiest to just get their output profile to make sure their results will be satisfactory. Generally, I produce the reproductions myself so it's not often an issue.

[Doug Gray]

I find some or your terminology confusing. In an attempt to take the issue apart a bit, I offer the following:

Conversion of raw image values to an input-referred XYZ form (or any other linear RGB color space) using a compromise matrix is linear using most adaptation algorithms.
Conversion of raw image values to an input-referred XYZ form using a 3DLUT is in general nonlinear.
Conversion of raw image values to an input-referred sRGB form using either a compromise matrix or a 3DLUT is nonlinear.
Conversion of input-referred XYZ values to output-referred XYZ values is in general nonlinear.
Conversion of output-referred XYZ values to output-referred XYZ values in the presence of gamut limitations, using Absolute Colorimetric, is linear for in gamut colors, and nonlinear for out of gamut colors.
Conversion of output-referred XYZ values to output-referred XYZ values in the presence of gamut limitations, using Relative Colorimetric, is possibly linear for in gamut colors, and nonlinear for out of gamut colors.
Conversion of output-referred XYZ values to output-referred XYZ values in the presence of gamut limitations, using Relative Colorimetric, is likely nonlinear for in gamut colors, and nonlinear for out of gamut colors.
Conversion of output-referred XYZ values in the presence of gamut limitations to colorants of a particular printer is in general nonlinear.

I've used XYZ as a standin for any linear RGB color space in the above. Does that make sense to you?

Jim

Sure. The simplified, linear, transforms would work great if the RGB filters were colorimetric (linearly convertible to CMFs) but, of course, they are not. 3D LUTs can be used to remap the filter errors to produce a more perceptually colorimetric result but can never perfectly render of course. I've done that in the past using CC CG and PM5 but these days prefer the 2 1/2 LUT approach using Adobe camera profiles.

One thing that would be fun to experiment with is imaging a scene twice with an additional lens filter to create, in essence, more than 3 color channels to improve the match to CMFs and keep the matrix process linear.


As an aside, in repro work I encounter gamut limitations infrequently. Largely because conversion to scene referenced does not involve the large increases in saturation typical in output referenced conversions.

[Jim Kasson]

One thing that would be fun to experiment with is imaging a scene twice with an additional lens filter to create, in essence, more than 3 color channels to improve the match to CMFs and keep the matrix process linear.

That's an excellent idea.

As an aside, in repro work I encounter gamut limitations infrequently. Largely because conversion to scene referenced does not involve the large increases in saturation typical in output referenced conversions.

I've found output referred saturation errors to be negative to modest with C1 and Adobe default profiles:

http://blog.kasson.com/?p=12867

Jim

[Doug Gray]

I've found output referred saturation errors to be negative to modest with C1 and Adobe default profiles:
http://blog.kasson.com/?p=12867

Jim

Indeed, your observation re colorchecker colors is right. Many colored patches are reduced in saturation when captured as output referred images in the PS RAW default. It varies somewhat with vendor. For instance the images created in camera in sRGB on my 5DS R are more saturated than the same image processed by Adobe Raw.

The main reason for this is that the colors that increase in saturation are those with RGB values in the midrange. Saturation increase is a byproduct of contrast increase which occurs in the midrange. High levels of R,G, or B are softly rounded and not hard clipped and this effect significantly reduces saturation. Many of the patches on a colorchecker have at least one channel that has a high enough value to be impacted. Whether the saturation of a particular color is increased or decreased depends on the whether the RGB values in the high contrast region outweigh the rounding that occurs at the higher levels.

Typically, Values in the range of RGB(100,100,100) are near the max in terms of contrast increase. You can measure the effect by printing the 9 colors around RGB(100,100,100) increasing and decreasing each channel by about 10. Also, print the 9 patches around RGB(220). Now take a picture of both of these in the same image. You will see an increase in saturation on the 8, dimmer non neutral perceptual intent patches v the 8 brighter patches. The brighter patches are less saturated than they would be with a scene referenced capture while the others are more saturated.