Bit Depth and Color Spaces

[tthaley]

As someone who regularly works with an 8-bit and 10-bit display (side by side), I would not agree with this. There is a difference.

This seems to contradict the following, from the bit depth tutorial:

The human eye can only discern about 10 million different colors, so saving an image in any more than 24 bpp is excessive if the only intended purpose is for viewing. On the other hand, images with more than 24 bpp are still quite useful since they hold up better under post-processing

[Manfred M]

This seems to contradict the following, from the bit depth tutorial:

Not at all, as 10 million colours defines the entire gamut that the human eye can see. This has nothing to do with how a display generates the different colours. which is driven by the bit rate of the display; a sRGB display will use a 8-bit per channel configuration (and can only display 35% of the colours that we can see) and a (almost) AdobeRGB display will use a 10-bits per channel display approach that lets it display about 50% of the colours a human can see.

[Mark von Kanel]

WHOOOOSH!! (he ducks as it all flew straight over his head) so what settings do i use?

[Donald]

WHOOOOSH!! (he ducks as it all flew straight over his head) so what settings do i use?

Good to have company. Science and me never did understand each other.

[Mark von Kanel]

Good to have company. Science and me never did understand each other.

As i get older it starts to get harder to cram it all in! all i want is a set of bomb proof settings to use to give best possible results, but i suppose it will never be that simple! so i guess ill look at the camera settings and duplicate it all from there.

Im hoping Manfred will crop up and say "use these settings" because my equipment i think is pretty much identical to his

[tthaley]

Unfortunately, it's pretty difficult to tell whether or not someone is giving good advice. "Professionals" have given me advice in the past, only to find a year later it was terrible advice. Learning all the underlying reasoning allows one to justify why doing something one way is better than another. (This is not to say anything negative about Manfred.)

[Mark von Kanel]

Yes i understand this tom, ill go a read some more and try and make sense of it, in the scheme of things im sure it doesnt really matter at my level. ill just shoot more B&W

[Manfred M]

WHOOOOSH!! (he ducks as it all flew straight over his head) so what settings do i use?

The "best" settings your equipment can support, unless there is an overhelming reason not to. That's what the design paramters for the equipment were designed to deliver, so why would you not use them? That's what you paid for in the first place.

[Manfred M]

Unfortunately, it's pretty difficult to tell whether or not someone is giving good advice. "Professionals" have given me advice in the past, only to find a year later it was terrible advice. Learning all the underlying reasoning allows one to justify why doing something one way is better than another. (This is not to say anything negative about Manfred.)

Tom - I don't claim to be an expert, but have found much the same as you have, I've received "misguided" advice from experts, so being a technical kind of person, and I’ve done the research on it so I can figure things out for myself. Part of the problem is that different aspects on the subject we are looking at use the same terminology for different things and at times there are connections and other times there are not.

When I first got into serious digital photography (I had been shooting film since I was in high school), I decided to get a bit of a refresher and enrolled in some photography courses at the local community college. One would think that the professors there, who were both very experienced professional photographers and went through the vetting process to become instructors in the photography program, so one would expect a very knowledgable person who knew the right answers. Frankly, some of the answers I got were "wrong" for the photography I do, but may have been right for them. The most blatent example is that most of them shot jpeg only and found RAW a waste of time; BUT (here is where the context comes in), a number of them came from a wedding photography background , where they were getting the shots to the point where their clients were happy with SOOC shots.

From a technical standpoint; one should always work with the highest quality data and not truncate it until the very last step. Once you lose precision by rounding, you can never get it back. If you apply a 8-bit operation to a 16-bit data field, you may still have 16 bits of data, but it is only accurate to 8 bits. That's a basic tenet of math.

Unfortunately, the field of photography and computers is muddled by marketing people. The terminology in much of the data we see about our gear is marketing baffle-gab, with no technical (or legal) meaning. You can be pretty sure that the manufacturer's legal department sets the guidelines and reviews the material to ensure that it can be defended in a court of law, if a legal challange is ever launched.

Much of the equipment we use is driven by proprietary software, so we can't just "peak under the hood" to see how it works. Same comment goes for the software that we use to manipulate the images.

Let me go into a few specific examples.

As an example, we talk about 16-bit a lot, but our cameras deliver, at best 14-bit images, so what's this 16 bit "stuff". Simply because computers, being digital devices work in 8, 16, 32 and 64 bit data, so 14-bit data gets 2 additional bits packed on the front to make things compatible with 16-bit computer data string size, etc.

The other issue is that there is really no "right" answer and there are all kinds of "yes, but..." answers involved. One has to understand these to make informed decisions. And of course, technology moves ahead and things that were true even a year ago, may not be so today...

A few truths I've dug up...

1. The human eye / brain can resolve around just under 10 million distinct colours / shades; but that is going to vary a lot from person to person and can be related to gender (women often have better colour acuity than men), age (colour acuity tends to go down as we age and there is some yellowing of some of the eye). This can be influenced by environment (impact is higher at high altitudes where eyes are more subjected to the damage caused by UV).

2. Colour spaces (gamuts) are mathematical models with specifically underlying technical issues that they are designed to resolve. When we say a 8-bit sRGB gamut can represent approximately 16 million distinct colours, that is technically corrected (each of the three primarly additives in 8-bits means that there are 256 distinct shades of pure red, 256 distinct shades of pure green and 256 distinct shades of pure yellow). Multiply that out = 16 million distinct colour combinations that can be described using 8-bit data.

As 16 million is significantly larger than 10 million, many people feel that sRGB must be able to display more than the human eye can resolve. This WRONG as the way sRGB was designed (to be able to describe what a computer screen was capable of displaying); in fact it covers approximately 35% of the 10 million distinct shades a human can see. Let's face it, digital photography was not a driver in the design of this colour space.

AdobeRGB was developed to unify (in RBG terminology) how digital data displayed on a screen could be descibed to a digital printer. We often hear that photo printers are RGB devices; THAT IS NOT CORRECT. They all all use a variant of a CMYK colour model (in the most basic printer); all of the cartridges in my colour laser printer are black, yellow, cyan and magenta. My pro photo printer is a bit more complex; it has blacks and grays, as well as multiple cyan and magenta cartridges and a single yellow. However, the software drivers map our RGB screen data so that it can be represented by a CMYK printer. That was the chief design concern when AdobeRGB was developed, and remember, Adobe has a whole suite of products for the graphics and publishing industries, so again photography was not the primary driver in this colour model. If covers just over 50% of the colours we humans can see,


Enter the wide gamut colour models, all attempts to descibe more colours mathematically. ProPhoto is simply one of many and has become popular because it appears to cover 100% of the colours we humans can see.

3. Computer screens (don't call them monitors, that is a completely different beast (very few of us would spend the money on one) are a product of design compromises; price being a very important one. There are very few companies in the world that actually made large display screens ( 3 or 4 at most), and these facilities produce screens used in televisions and computer screens. The underlying technology is similar, but there is a world of difference in producing a screen that produces a flicker free, punchy image that is required to watch sports (frame rates vary from 24 to 60 frames per second), versus a static image for editing. Supporting these frame rates means some design trade offs have to be made, and the one that affects photographers the most is colour accuracy. Just because a basic screen can support the standard sRGB colour space, the technology to do so has NOTHING to do with how the screen reproduces the colour. That is related to the technology used. A decent explanation is:

http://en.wikipedia.org/wiki/LCD_monitor

So, bottom line is that so far as I am aware, sRGB screens come in two flavours; 6+2 bit (where the screen has a native 6-bit resolutio and the other 2 bits of data are handled by the screen drivers dithering (nice term for turning the pixels off and on rapidly) to give us the 16 million colours. Higher end technology is true native 8-bit, and dithering is used to give us the 2 additional bits for 10-bits per channel. The issue is, of course, being proprietary, we have no idea as to how accurately any of these displays show the sRGB or AdobeRGB colour space. Profiling and calibration tweak the performance for improved accuracy, but honestly, the two calibrated and profiled screens I am looking at right now reproduce colours differently.

4. While we can only see a subset of the colours that our cameras have captured on our computer screens, what is important to us in editing is to ensure that we understand the impacts of our edits. Changing even some very basic parameters (a small change in exposure, for instance) means that the specific pixels are displayed on our computer screens (remember, we don’t necessarily care what is happening, what we are interested in is ensuring we get a good looking image. What we care about is how the computer screen (or in the case of a print, the printer) creates the output and how our eyes / brain interpret it.

With a smaller colour space, like sRGB, the discrete steps between two colours are fairly small, so even an extreme edit may not have an impact. I’ve done lots of very successful edits on jpegs without any noticeable artifacts. So if push comes to shove, working in 8 bits on sRGB is probably going to be fairly safe. Taking the stair analogy, if your steps are 5cm / 2-1/2” apart, removing every other step is not going to make using the stairs significantly more difficult.

The same cannot be said for a wide gamut colour space like ProPhoto. The steps between two colour shades are fairly large, in absolute terms, and if you are working in 8-bit mode, there is a higher likelihood that you are going to create a visible artifact. With ProPhoto, 16-bit is probably a much safer option than 8-bit. Back to the stair analogy; if your steps are 15cm / 6” apart and you take out every other step, climbing stairs with a 30cm / 12” rise is going to be much more noticeable.

[pnodrog]

The "best" settings your equipment can support, unless there is an overhelming reason not to. That's what the design paramters for the equipment were designed to deliver, so why would you not use them? That's what you paid for in the first place.

Yes it is just that simple.....

[Colin Southern]

As an example, we talk about 16-bit a lot, but our camerass deliver, at best 14-bit images, so what's this 16 bit "stuff". Simply because computers, being digital devices work in 8, 16, 32 and 64 bit data, so 14-bit data gets 2 additional bits packed on to make things compatible with 16-bit computer data string size, etc.

Fun fact (which I think still applies).

Photoshop actually only uses 15 bits in "16 bit mode".

Have a read through all of these posts (several times

http://www.photography-forums.com/ph...de-t70203.html

[Colin Southern]

Yes it is just that simple.....

Good grief man - you can't just go around making real-world common sense statements like that

10, 9, 8, ...

[ajohnw]

Fun fact (which I think still applies).

Photoshop actually only uses 15 bits in "16 bit mode".

Have a read through all of these posts (several times

http://www.photography-forums.com/ph...de-t70203.html

Are you sure Colin? I doubt it some how but maybe they did at some point. They could as far as what might be called output is concerned but when it comes to manipulating gamuts that would fall flat on it's face as accuracy would be lost at a rate of knots. When we look at the stuff we are looking at integers. I've yet to hear of a floating point monitor.

On the rest I feel people really need to think "fractional bits" when considering workspaces that are greater than the one that is being "used" and when info is greater than output.

I just gave myself an update on current display technology. I already knew that even Linux will handle more than 24bit colour. I'm fed up of typing 3x8bit so will use computing words. It seems HDMI will handle 36bit but may have bandwidth problems on larger displays. Displayport can handle 30bit on larger display or daisy chained smaller lower resolution displays.

Ok so some one is viewing and adjusting a wonderful 30bit image on a true 30bit display that may even have the same granularity as sRGB and have achieved this incredible result. Only problem is that it's not sRGB and according to Adobe's own specification not aRGB either neither is it prophoto so what are they going to do with it? As soon as they select a common gamut it will change. The only aRGB spec I have seen is on Adobe's web site and only mentions 24 bit colour. The wiki entry also mentions the colour gaps in it because of this. It's an interesting read in relationship to mistakes too.

The ViewSonic monitor mentioned is amazing. Info on that shows 36bit input processed internally at a higher bit depth and then shown on a 30bit panel - true ones of those are still thin on the ground. The word dithering gives me the shivers from ealry colour PC monitors. Probably shouldn't on these panels but I have a natural aversion to anything that uses it.

ProPhoto workspaces - that is really weird. The wiki shows the maths behind all gamuts that is similar to converting from one to the other. it seems data goes

Camera -> Camera Profile -> ProPhoto -> what's actually needed except a prophoto printer which in real terms probably has it's own gamut that doesn't really cover it anyway plus the fact it's a bit difficult to print imaginary colours. The choice of those is interesting too.

The camera profile is needed but I can't see the point in translating that to a true prophoto numeric format when in all cases it's going to get translated to something else in order to make any use of it. It would be unusual for lower level software people to do something so inefficient so I can't help wondering if the numbers are just stuck in a 48bit colour space and then processed to what is needed. It even has a D50 white point. The common gamuts use D65. I wonder if it's all another case of hiding binary bits from the public really and they really do leave it as is converted and temperature corrected via the camera profile.

The problem with gamut translation is the 3 fully saturated colours at the ends of the triangles usually shown. Simply put the blue and green numbers used with zero red move the colour along that side of the triangle between the blue and green the gamut chooses to use. That aspect sets the over all size of the gamut. 24 bit colour could cover the prophoto gamut but the steps would be a bit big. Why map camera colours to some colours that don't even exist - the two imaginary ones prophoto uses only to go through another set of translations to some other gamut. Maybe they do but I have to have my doubts.

John
-

[Mark von Kanel]

The "best" settings your equipment can support, unless there is an overhelming reason not to. That's what the design paramters for the equipment were designed to deliver, so why would you not use them? That's what you paid for in the first place.

so as i said pick the best colour space settings on the camera with all the bits you can pack in and then set the rest of the equipment to match simples

[Manfred M]

Fun fact (which I think still applies).

Photoshop actually only uses 15 bits in "16 bit mode".

Have a read through all of these posts (several times

http://www.photography-forums.com/ph...de-t70203.html

Which is fine, because camera output is only 14-bit. My understanding is that even medium format camera that produce 16-bit output are creating 14-bit data that is packed with a couple of zeros.

Using 15 bits makes a lot of technical sense, as the remaining bit is used to denote whether we are dealing with a positive or negative number.

[ajohnw]

Which is fine, because camera output is only 14-bit. My understanding is that even medium format camera that produce 16-bit output are creating 14-bit data that is packed with a couple of zeros.

Using 15 bits makes a lot of technical sense, as the remaining bit is used to denote whether we are dealing with a positive or negative number.

LOL I would love to see a gamut conversion done with signed 16 bit.

John
-

[Manfred M]

LOL I would love to see a gamut conversion done with signed 16 bit.

John
-

I wouldn't be too surprised if this is used internally when running some of the blending mode effects and speciality filters.

For basic image manipulation, of course not. I just remember writing some image manipulation algorithms years ago where I used this approach (used the roll-over as part of an edge detection approach). The output was fed into a neural network package to identify overlapping objects in an image and fed that into a data analysis system. Probably the last bit of production software I wrote.

[Mark von Kanel]

Well i checked my camera and its set to sRGB, so now what? if i change it to Adobe i know my high end NEC monitor (yes Manfred it is a monitor ) at home will display this gammut but can i configure my macbook pro (run in windows) matte screen to display adobe, ive had a quick dive through the settings and it would appear that it can, anybody done this? it has settings for 32bit true colour and 16bit high colour, but as these are windows configurations i dont know if the screen will display them, ill have to google it!

And if i alter all these settings will it have any visible effect when i print? im certain it wont make any difference to the average users viewing on their PC!!

But then again if i process in the highest colour space i can then im less likely to get artifacts... yes? so it will be worth it in the end?

[Manfred M]

Well i checked my camera and its set to sRGB, so now what? if i change it to Adobe i know my high end NEC monitor (yes Manfred it is a monitor ) at home will display this gammut but can i configure my macbook pro (run in windows) matte screen to display adobe, ive had a quick dive through the settings and it would appear that it can, anybody done this? it has settings for 32bit true colour and 16bit high colour, but as these are windows configurations i dont know if the screen will display them, ill have to google it!

And if i alter all these settings will it have any visible effect when i print? im certain it wont make any difference to the average users viewing on their PC!!

But then again if i process in the highest colour space i can then im less likely to get artifacts... yes? so it will be worth it in the end?

Mark - I'm not normally a Mac user and most of the time I use one (a Mac Pro, connected to a MONITOR), it's at the video co-op I belong to, so I'm on it doiing video editing and not paying attention to the settings. Try it, and if the screen display works, you should be fine; if it looks like crap, then go back to your original settings. The newer MacBook Pros all have higher end IPS screens, so you paid for 32-bit TrueColor, so why wouldn't you use it?

If you are posting on the internet, you should output to sRGB jpegs regardless, as this is what the colour-managed browsers can handle, if you don't downsample to this colour space, the posted images could look muddy.

The only time I've run into serious artifact issues is when I apply heavy duty corrections to a sub-optimally exposed shot; and here the issue usually shows up as blocking in in the sky; dialing things back a bit of more serious surgery (manually introducing a better sky shot) will usually solve the problem.

If your camera is set to sRGB, it will only affect the jpegs anyways, and if you are editing RAW, the colour space is only assigned when you do the RAW conversion.

[Mark von Kanel]

thanks Manfred, thats really helpful, i guess i should have realised that the in camera setting would only effect JPEGS DOH!! (i seem to be saying a lot of that lately ) i only shoot raw so its all moot really because i can set whichever colour space i want to output in, in lightroom or PS.....

and then when i export i can do it in whatever colour space is suitable Which will mostly be sRGB?? ive just wasted a load of time thinking about something i can change backwards and forwards as much as i like... i think....

But what about printing, isnt this gamut lark just a monitor (sorry display screen ) thing, will it make any difference which space you work in for printing?

I will get the hang of this PP lark eventually!