Pages: [1]   Bottom of Page
 Author Topic: How do out of gamut colours print using relative colorimetric?  (Read 3570 times)
JohnCoveney
Newbie

Offline

Posts: 5

 « on: March 15, 2012, 06:25:42 AM » Reply

Only my second LuLa post (first was in 2009) so not sure if I should post this question here or or in the Colour Mgmt board.

Anyway, I've been working my way thru the camera to print and screen tutorial and in the discussion of perceptual vs. rendering intent at the start of segment 23, Schewe says that out of gamut colours are clipped when using relative colorimetric. Perhaps someone could give a bit more detail as to what actually happens - does the printer (an Epson 3800 in my case) simply pick the nearest in gamut colour and use that? By clipped, does that mean the chosen colour will be fully saturated - if that's even a sensible question?

Sorry, if this has been dealt with already but I couldn't find it - if it has, a link wd be great

Thanks, John C
 Logged
michael
Sr. Member

Offline

Posts: 4731

Here's a simple analogy that I use to explain the difference between Relative and Perceptual. Imagine that your wife has asked you to put a pillow in a pillow case. It's a big fluffy down pillow.

You can two choices. 1; stuff the pillow into the pillow case. 2; cut away the parts of the pillow that don't fit. The first is Perceptual the second is Relative.

There's more to it that than, and I'm sure that someone will chime in quickly with an in-depth technically accurate description, but I've got to run, and the pillow analogy is a good first approximation.

Michael
 Logged
mac_paolo
Sr. Member

Offline

Posts: 418

simply pick the nearest in gamut colour and use that? By clipped, does that mean the chosen colour will be fully saturated
As for what I understood so far, this is the right answer.
Relative colorimetric is better for those images with small over-saturated peaks here and there, without out-of-gamut smooth gradations.
Perceptual actually forces the whole image in under saturated feel because of those small peaks.
 Logged
digitaldog
Sr. Member

Offline

Posts: 8091

Sometimes, a picture is worth 1000 words so this illustration may help:

RelCol uses a Gamut Clipping (what Schewe likes to call the sex change operation). Perceptual uses a Gamut Compression. Neither is better or worse, the choice should be made after viewing the two and picking the process which produces the appearance you prefer.

The other very important key to understand is that there is no standard way to conduct a Perceptual rendering. Every profile making software can come up with a method to do this. Just like when viewing E6 film, Velvia looks different than Ektachrome. The manufacturers of those films, like the people who build Perceptual renderings produce what they think their customers may prefer. So the quality or acceptability of a Perceptual rendering can vary. A lot! Non out of gamut colors can be affected with the Perceptual rendering because the idea is, alter the out of gamut colors to maintain a relationship with other colors so you perceptually deal with this process in a (better?) fashion.

There’s more to all this of course. For example, how a rendering intent treats an image that began in sRGB versus one that began as ProPhoto (by the time the rendering intent comes into play, the CMS has no idea which was which).

I’m wouldn’t dare ask Michael about pillow biting after that last post <g>!
 Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
JohnCoveney
Newbie

Offline

Posts: 5

Thanks for the replies, especially Andrew's - its clear to me now. I don't need to worry about how the image started as I shoot in RAW & process in Prophoto.

John C

PS: Given MR's & JS's analogies, I sometimes wonder what they get up to when Chris turns off the camera!
But I don't need an answer to that!!
 Logged
Rhossydd
Sr. Member

Offline

Posts: 1862

.....RelCol uses a Gamut Clipping......
My understanding of this is that it's a little bit more complex than your illustration suggests.
I've read that in RC as one approaches the gamut boundary a degree of colour compression happens to prevent everything just clipping completely.
To use the pillow analogy; everything thing through most of the pillow stays exactly where it was, but the feathers at the outer edges get pushed tighter together, but don't actually overlap.
 Logged
digitaldog
Sr. Member

Offline

Posts: 8091

My understanding of this is that it's a little bit more complex than your illustration suggests.

It is actually an illustration done by Karl Lang. I don’t think either of us want to go there (if you know Karl) <g>.
 Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
bjanes
Sr. Member

Offline

Posts: 2718

My understanding of this is that it's a little bit more complex than your illustration suggests.
I've read that in RC as one approaches the gamut boundary a degree of colour compression happens to prevent everything just clipping completely.
To use the pillow analogy; everything thing through most of the pillow stays exactly where it was, but the feathers at the outer edges get pushed tighter together, but don't actually overlap.

As far as I know, Karl's diagram for colorimetric perceptual rendering is correct: in gamut colors are not affected and out of gamut colors are clipped without compression. The perceptual rendering illustration does not tell the entire story. With perceptual rendering, the color management module does not look at colors that are actually present in the image, but it merely applies a predefined amount of compression according to the profile. If the image has no out of gamut colors, it will still be compressed. On the other hand, colors that are far out of gamut may not be compressed sufficiently to bring them into gamut. As the DigitalDog points out, the degree of compression is dependent on the maker of the profile. Perceptual rendering can reduce banding as with blue skies, but if no important textural detail is lost, it is often best to merely let out of gamut colors blow out.

This old post by Mike Chaney offers some useful explanation. That post is rather old and does not take into account changes made with version 4 profiles. However, version 4  profiles are not in wide use, and perceptual rendering is still problematic.

Regards,

Bill

Edited to correct a typo
 « Last Edit: March 15, 2012, 11:10:49 PM by bjanes » Logged
Schewe
Sr. Member

Offline

Posts: 5256

As far as I know, Karl's diagram for perceptual rendering is correct: in gamut colors are not affected and out of gamut colors are clipped without compression.

Uh, is that a typo? Shouldn't it read "As far as I know, Karl's diagram for relative colorimetric is correct: in gamut colors are not affected and out of gamut colors are clipped without compression."?
 Logged
hjulenissen
Sr. Member

Offline

Posts: 1615

So this is somewhat similar to regular audio clipping vs "soft clipping"? The first would preserve perfectly all samples that can be represented, while mapping those that cannot be represented to the closest legal value. The second would actually alter some legal values into other legal values, but in doing so could potentially preserve local waveshape "better" (prioritizing smoothness).

-h
 Logged
Rhossydd
Sr. Member

Offline

Posts: 1862

So this is somewhat similar to regular audio clipping vs "soft clipping"?
Maybe an audio analogy is that perceptual turns down the volume to fit, but RC puts a limiter in circuit.
 Logged
bjanes
Sr. Member

Offline

Posts: 2718

Uh, is that a typo? Shouldn't it read "As far as I know, Karl's diagram for relative colorimetric is correct: in gamut colors are not affected and out of gamut colors are clipped without compression."?

Jeff, It was a typo or mis-statement, which I corrected. Thanks for pointing it out.

Regards,

Bill
 Logged
hjulenissen
Sr. Member

Offline

Posts: 1615

Maybe an audio analogy is that perceptual turns down the volume to fit, but RC puts a limiter in circuit.
Turning down the volume globally would correspond to globally decreasing saturation until the entire image was within gamut. My understanding is that this is not done in either of the color mapping modes?

1. Clip only samples/pixels that cannot be represented
2a. Do a linear/non-linear mapping on a per-pixel basis such that extreme o-o-g pixels are brought to the clipping point, while moderate o-o-g (and saturated pixels within gamut?) are moved further towards white.
2b. Alter a (more or less) local neighborhood of samples surrounding out-of-gamut ones, such that the local waveform is kept smooth

-h
 Logged
Rhossydd
Sr. Member

Offline

Posts: 1862

Turning down the volume globally would correspond to globally decreasing saturation until the entire image was within gamut. My understanding is that this is not done in either of the color mapping modes?
Actually for a page 1 explanation of the difference between PC & RC I think that analogy serves reasonably well.

As Andrew points out earlier in this thread, the nitty gritty of how perceptual rendering is handled varies between different manufacturers of profiling software and some even offer different choices too. Once you start reading up on this a masters degree in imaging science really helps too.
Colour management is a tricky business. At first encounter it seems to be an empirical science that solves all the problems of delivering "accurate" colour, but the deeper you delve into the technicalities the more you discover that at some point it's just someone's idea of what is best and it has a degree of  subjectivity.
 Logged
 Pages: [1]   Top of Page