Color space. How to get lost in space

[kontrol]

DISCLAIMER : You might not understand what I wrote because I'm not pretty good with that subject and probably mixed many concepts all together .

So, yeah Color space.
I read so much about it, here in the tutorials, tips on the web, and now this book (Photoshop for Photography) I'm reading.
I did read the color space tutorials here and I got lost.

So let me first say how I'm currently setup.

My Camera is a Canon T3, in the settings I use sRGB. Adobe RGB is the other setting available. All my shots are in raw.
In lightroom 3, I think my photo opens in sRGB.
In lightroom 3, when I right-click->Edit in photoshop, in the settings, the picture are opening in Photopro RGB?
I save my files as TIFF, and I'm not exaclty sure on what color space again. (The one it is opened?).

So here's a list of question I keep asking myself about all this so I might finaly understand :

1. What's up with color space. I have a Canon T3, so 12 bits per channel for my raw files, which mean 4096 possible value for each color. Now how opening, or taking the photo in a different color space would "change" this value. The value is fixed, if it is 0, 4096, 0, it has to be green, not red right?. Taking the photo directly to Adobe RGB instead of sRGB won't change the pixel value right? Is it only to put the default color space so the photo is opened in the right color space?

2. Opening a photo in a color space. What does it mean? Is a color space an "Interpretation" of the pixel value for my screen? A photo is taken in sRGB, opening one in sRGB and one in Adobe RGB might not show the same on the screen?

2.1 sRGB for lightroom, photoshop and every other application mean the same color?

3. What is the ideal set up? Should i put my camera, lightroom and photoshop default to the same Color space?

4. Should I use Adobe RGB or sRGB? I see Adobe RGB have a larger gamut of color, but I'm not even sure how I understand that. I might understand I understand your answers to my previous questions .

5. Printer also have Color spaces? Is it their own interpretation of the Pixel value too?

[nickjohnson]

Hello Mathieu.
You ask some very fundamental questions – it took me forever to figure this stuff out. Here are some answers.

Background.
All physical devices – camera, screen, printer – have their own proprietorial device colour space. Essentially these are prescribed by the physical characteristics of the device. They are a closely guarded secret of the manufactures. You can do nothing with them – quit worrying about them.
The various colour spaces that you mention – sRGB, Adobe RGB, ProPhoto – are all container colour spaces used to transport data to and from physical devices and for manipulation of images in – for example – Lightroom.
Camera colour space settings only impact jpeg production – including the pic on the back of your camera. Raw files have the jpeg in them – but have no colour space assigned since the data is Raw.
To the best of my knowledge sRGB is a lowest common denominator type colour space. Adobe RGB is a super set of sRGB. That is – all of sRGB fits inside Adobe RGB. ProPhoto is a superset of Adobe RGB, and is the widest of the three colour spaces in common use.

With that all in mind here are the answers to your questions.

1.The choice of colour space has no impact on pixel values. The choice does impact the fidelity or graduation between different values, and also the actual colour used or perceived since some colour spaces are bigger than others. It is inevitable that your camera will record data at the sensor and in a Raw file that will represent a colour that is outside some or all of the colour spaces.

2.A colour space is an interpretation of the incoming data stream. If the interpretation is of Raw data by a Raw converter – the colour space is assigned to a saved file – e.g. a tiff file. If the incoming data stream already has an assigned colour space – e.g. tagged as sRGB, an application like Lightroom may well ask about opening the file in a particular colour space, if the working colour space of the application is set to something other than that of the incoming data stream. This situation is a colour space mis-match so the application will typically ask what to do to resolve a mis-match. As standard Lightroom v3 uses ProPhoto.

2.1Yes it should (but may not)!

3.If you wish to use jpeg exclusively that is a good plan. If you wish to get the best quality save Raw files (or Raw + jpeg) then set LR and PS to ProPhoto

4.See the above.

5.Yes they do. However they interpret the incoming data stream which is – in practical reality – the values of a colour space and not the pixel values recorded by the camera.

HTH

Regards,

Nick.

[Colin Southern]

3.If you wish to use jpeg exclusively that is a good plan. If you wish to get the best quality save Raw files (or Raw + jpeg) then set LR and PS to ProPhoto

Hi Nick,

Nice write-up

One thing that I like to point out to people is that the bigger the colourspace, the bigger the trouble people can get themselves into if they don't know what they're doing; and saving photos in wide spaces like Prophoto are a good example. Yes, it can hold more information, but the portions outside the sRGB gamut can't be seen on many monitors because they're essentially only sRGB devices - so folks are essentially "working blind" (or worse, in that the colours they THINK they're getting may not be what they're getting come print time).

Printing can also be an issue with wide spaces, as the vast majority of print shops only handle sRGB, and if one forgets to convert it (for print or Internet display) then the results can look pretty bad (especially portraiture in Prophoto that's assumed to be sRGB -- don't ask me how I know this!).

So often I suggest to people that they stick with sRGB (I think of it as SAFE RGB) unless they're doing their own printing, or having it done at a high-end print shop - there's usually little lost in terms of colours (most wouldn't notice anything), but the changes to levels / saturation etc can be quite pronounced.

[nickjohnson]

Hello Colin,

First up, I do hope that all is well with you and yours. The news footage coming out of Christchurch today was hart rending.

I take your point about the limitations of screens when used to edit wide colour spaces like ProPhoto. I would point out that that is why the perceptual rendering intent was invented. More problematic are printer / ink / paper combos that can print darker blacks than most practical monitor set-ups can display. Still – it all adds to the fun.

Without getting into the nuts and bolts of jpeg, it must be kept in mind that it is engineered and intended as a final output file format. That means that ANY and ALL editing – no matter how it is done is likely (some would say it's inevitable) to reduce quality and make consistency difficult. Thus my reference to Raw + jpeg. I know that doing that will increase storage costs and cause buffer problems with action shooting. The upside is that when / if one revisits past work with some new, hard one pp skills, one has the best quality data available.

Ah – compromises, compromises.

[kontrol]

So if I understand,

Color space has nothing to do with raw files? Which mean that my photo register up to PhotoPro RGB? And It's up to me to keep that or work in a smaller color space?
So it doesn't matter the setting in my camera. What change is the setting in Lightroom when I import them?

To show the raw picture on the screen, Lightroom has to use a color space right?

I'm best using the same color space from the camera to photoshop?

Also when I save a file in a color space, I change the pixel values?

Lastly I understand that Monitor doesn't show more than the sRGB gamut, so what's the point of working with anything greater than that if you can't see the difference. In the end you might end up with prints that might not look like your screen right?

EDIT : Maybe I got it wrong again (heh). By re-reading your answer, You say color space doesn't change the RGB value. Does it mean that in a sRGB color space that two different pixel value could show up as the same color because the gamut is not big enough to show a difference between the two?

[nickjohnson]

Mathieu,

Color space has nothing to do with raw files?

Raw files have no colour space in the sense of sRGB, etc. Thay have a device colour space that is the result of the physical characteristics of the camera. When you feed the raw file into a converter – like Lightroom – the converter assigns a colour space as part of the conversion process.

Which mean that my photo register up to PhotoPro RGB? And It's up to me to keep that or work in a smaller color space?

Actually your camera Raw files may or may not present data that fits inside a particular colour space. It's the job of the raw converter – like Lightroom – to take care of this for you.

So it doesn't matter the setting in my camera. What change is the setting in Lightroom when I import them?

For Raw data that is correct.

To show the raw picture on the screen, Lightroom has to use a color space right?

Yes, and Lightroom passes it's colour space data to your computer operating system's graphics sub-system – where the data is automatically translated into the colour space of the physical output device - the screen that you look at.

I'm best using the same color space from the camera to photoshop?

Please see answer 3 in my original post - above.

Also when I save a file in a color space, I change the pixel values?

That depends upon many things – but mainly upon what space your coming from, what space your going to and if the data value is interpreted the same or differently if the spaces are not the same.

Lastly I understand that Monitor doesn't show more than the sRGB gamut, so what's the point of working with anything greater than that if you can't see the difference.

When sRGB was invented in 1996 it was intended to match the common capabilities of output devices such as screens and printers – at that time. Modern screens can display more than sRGB and some very expensive ones can display Adobe RGB. Modern photo printers – including domestic models - can exceed Adobe RGB in some photographically significant areas. The problem with using sRGB as a colour space for editing images is that it uses 8 bit encoding for each channel – Red, Green, Blue. Adobe RGB and ProPhoto RGB typically use 16 bit encoding (some apps use 32 bit). Whilst the 8 v 16 bit comparison is difficult and tedious – not to mention non linier – the fact is Lightroom uses ProPhoto as the installed default for a reason. The reason is that 8 bit encoding has insufficient resolution to make a good job of coping with the image manipulation that modern applications are built to achieve. Those who wish to argue this point are respectfully requested to do so with Adobe, Apple, whoever.

In the end you might end up with prints that might not look like your screen right?

By default it is highly unlikely that you will see a good correlation between your screen and your printed output. However – a useful and informative correlation is possible and achievable by means of controlled viewing conditions, properly profiled printer / ink / paper, a calibrated screen, and a fully colour managed workflow.

Does it mean that in a sRGB color space that two different pixel value could show up as the same color because the gamut is not big enough to show a difference between the two?

In any colour space two different data values can end up as the same colour. In that case the cause may be insufficient encoding resolution to distinguish between them. Where the data values fall outside the colour space of the display device the colour actually shown may be the same as a result of the automatic colour management going on in the operating system and also any rendering intent settings.

[Colin Southern]

Lastly I understand that Monitor doesn't show more than the sRGB gamut, so what's the point of working with anything greater than that if you can't see the difference.

Many modern monitors are starting to inch past the sRGB gamut, but there's a couple of things that I can think of that you might like to keep in mind ...

1. Even if you have a $7000 Adobe RGB capable Eizo monitor, it doesn't mean that others will have also, and it doesn't mean that they'll be running a colour-managed browser that'll convert the wide gamut file to something that their screens can understand -- so you run a significant risk that folks won't see your image like you do.

2. Printers can easily exceed some areas of the sRGB gamut.

So basically - if you're putting it on the net, stick it in sRGB - if you're having it printed at an off-the-street photo shop, stick it in sRGB - but if you're printing your own - or won't be displaying it publically - and are running a profiled & calibrated monitor and colour management aware software, then you may as well keep in in something wider.

[kontrol]

You guys are awsome

I re-read the Color space article, and the other in the color management section that I did not see, and I understand a lot more now the color space and conversions. + All your explanation, I might not understand 100% but still enough to be fine with it .

I still have 2 remaining questions on the subject.

1. Since sRGB is 8bit per channel, I should set all my devices to Adobe RGB (16bit per channel) because the applications recommand 16 bit per channel, and also to avoid posterization (as explaining in one of the many wonderfull tutorial here)? Even if my screen doesn't show it completly.

2. So just to make sure I understood the color space conversion, if I take an image and save it in loop between 3 (enough) different color space, I might end up with different color (talking about extreme color of the gamut) all that is of course theorical, and probably not really possible. I guess

[herbert]

Hi Mathieu,

Working in 16 bit will be nice if you are pushing the colours around. Imagine moving an object 45 cm but only having a ruler with intervals every meter and you can only place the object on an interval. Things will be put in roughly the right place but it will not look correct. You need a ruler with intervals every millimetre.

16 bit files are bigger though with modern computers the storage space is not much of an issue any more. Also in photoshop some filters and plugins only work on 8 bit images. In this case you can convert the image to 8 bit if you need to.

Question 2 is more important to understand. If you have colours that can fit into both colour spaces then when you convert between them you wil not have any colour changes. You can also convert back without loss. There may be some tiny differences due to rounding in the conversion but your eyes will not notice.

If you have colours that fit into one space and not the other then you have a problem. My analogy is that you are trying to fit a square peg into a round hole. However the square is too big. So what do you do?

There are two solutions. One is to squeeze all the colours together a bit so they all fit into the new space. The distances between them will change but the gradations of tone should be preserved. This is called perceptual rendering because your eyes will perceive the image appearing the same. This is like make the square smaller so it fits in the hole.

The second is to match up all the colours that fit. Any colours that do not fit are set to the nearest colour. In this case several colours in the original space could end up the same colour. This is called relative rendering because all the same colours are set at the same distance relative to each other. However some have changed. This is because you have taken the square peg and chopped off the corners so it fits in the round hole.

When you want to go back from the smaller space into the bigger space your image will to match the original. This is because the colours where mapped and altered in the conversion. You do not know how to do the exact reverse mapping because you do not have all the information. In the examples above you either do not know how much bigger to make the square peg that was reduced, or you cannot add back the corners that were chopped off since you do not know where to put them (or that are missing).

Hope this helps,

Alex

[Simon Garrett]

Just one comment: "Color space has nothing to do with raw files?" All images have a colour space. The colour space is the frame of reference, or the units you are using. Like measuring distance: might be feet, metres, nautical miles or whatever.

sRGB and Adobe RGB are standard colour spaces. However, the data coming out of camera sensors isn't in a standard space, it's in units unique to that sensor. If you shoot jpeg, the camera converts to sRGB or Adobe RGB. If you shoot raw, the image data is left in the sensor units, and the raw convertor (or Lightroom, which includes a raw convertor) converts the units.

Colour management is not entirely logical or consistent; when I groped my way to understanding it I wrote a cheat-sheet at http://www.simongarrett.co.uk/Colour...CheatSheet.htm. I not suggesting it's better than other tutorials, but sometimes reading the story more than once helps complete the picture. At the bottom I've included links to quite a few other sites I've found helpful.

[Colin Southern]

Hello Colin,

First up, I do hope that all is well with you and yours. The news footage coming out of Christchurch today was hart rending.

Thanks Nick,

I've only felt one of the shocks (I'm 400km away) - but we've had our own flooding of recent. What a year

I take your point about the limitations of screens when used to edit wide colour spaces like ProPhoto. I would point out that that is why the perceptual rendering intent was invented. More problematic are printer / ink / paper combos that can print darker blacks than most practical monitor set-ups can display. Still – it all adds to the fun.

Please don't use the words "problem" and "printer" in the same sentence anywhere near me - I had one heck of a wrestling match just before christmas. Basic problem is that black ink on unsprayed canvas has an L value of 25!!! - even after spraying it only dropped to 17. I had to whack in a truckload of compensation (right to the max) in the profile, and even now it's not perfect - at least it's working again though. Didn't help that I had a roll of canvas with a faulty coating over part of it's width that grossly affected the black point as well!

Without getting into the nuts and bolts of jpeg, it must be kept in mind that it is engineered and intended as a final output file format.

I'd even go so far as to say ONLY a final output format - but I also think it does that very very well. I do think people get caught up in the theory -v- the practise a lot with it though. In theory, one wouldn't want to edit jpegs, but in practice - although big levels changes are certainly problematic, other forms of editing (say, cloning etc) are generally tolerated quite well. Personally, I just don't really see a down-side to just following best practice and storing in a full-range non-lossy format.