Fwd: Monitor calibration/profiling for video applications
My take on this is that the video guys 'hard-tune' devices to a specific primary/gamma set (to use terms loosely) whereas the ICC approach has a much higher degree of flexibility in that each device is profiled and each piece of media carries with it a profile that, in effect, contains expectations as to how that content should look. Until very recently the video guys didn't even have a standard for the display end of things - the electro-optical transfer function or EOTF - even though they have had for some time various standards for the recording side of things (OETF) which of course matter much less as the look of video is more often than not processed "for effect" (subtle or strong). Until ITU-R BT.1886 there wasn't a standard for end-user display. It would seem much more sensible to me for the video industry to move to an ICC-profile type setup where recording devices are profiled (perhaps less important because of post processing), displays are profiled and each video (however processed for effect) comes with a profile that aids a CMM to deliver the correct look on any display.
Is this a fair view of the world? (says someone who has just profiled his computer monitor with i1Profiler while reading Poynton's "Digital Video and HD" and about to recalibrate my plasma display with HCFR)
From: Dennis Dunbar <dennis@dunbardigital.com> Subject: Monitor calibration/profiling for video applications Date: 14 August 2012 21:03:44 GMT+01:00 To: colorsync-users@lists.apple.com
Over the last year or so I've been focusing on learning more and more about working on the video side of the industry (after 20 years as a high end retoucher) and have noticed their approach to color management is markedly different than the stills/print side I've worked in for so many years.
It seems on the video side they rely much more on LUTs instead of ICC profiles and argue that their approach delivers a consistency and accuracy un-achievable through the ICC approach. Here are a couple of links to software products that seem to be adopting this approach:
http://studio.spectracal.com/calmanstudio.html
http://www.lightillusion.com/lightspace.html
In the case of Lightillusion's Lightspace product I had a brief exchange with the head guy on their forums in which he insisted their approach was vastly superior to the ICC approach I am more familiar with, yet he failed to offer any specifics that really explained why to me.
Is anyone else familiar with the different approaches to color management? And if so could you help clear up the fog for me? (Do they really have something there, or are they just blowing smoke?)
Thanks!
Dennis Dunbar
From: Graeme Gill <graeme2@argyllcms.com> Subject: Re: Monitor calibration/profiling for video applications Date: 16 August 2012 03:30:55 GMT+01:00 To: ColorSync <colorsync-users@lists.apple.com> Reply-To: graeme@argyllcms.com
Dennis Dunbar wrote:
Is anyone else familiar with the different approaches to color management? And if so could you help clear up the fog for me? (Do they really have something there, or are they just blowing smoke?)
Mostly blowing smoke :-) Typically Video and Film are aiming at a very specific sub set of general color management: Emulation of a particular standard. The print equivalent is side by side proofing. So most of the talk is about calibration (ie. making a display device behave in a specified way). ICC profiling is more general and flexible, and provides a mechanism for actually achieving the color management sub set desired by video or film.
Another key difference between Video/Film and general color management is the mechanics used to transform color. Computer systems can use software, which is very flexible but not necessarily real time. Video/Film often have either very simple hardware controls (ie. "brightness" and "contrast" knobs. "RGBCMY" primary controls that don't work so well, per channel curves etc.) or the high end actually has hardware capable of per channel, matrix and 3D cLUT transforms (ie. all the machinery ICC profiles use). What's loaded into the 3D cLUT is the equivalent of a device link. (Note though that real time computer color CMM's can be implemented using GPU's.)
Graeme Gill.
From: THOMAS A LIANZA <tlianza@xrite.com> Subject: Re: Monitor calibration/profiling for video applications Date: 16 August 2012 12:07:27 GMT+01:00 To: graeme@argyllcms.com, ColorSync <colorsync-users@lists.apple.com>
Hi to all,
Graeme's comments, as usual, are spot on. I work with the ICC and have worked with the Academy on issues of color management, so I think that I can speak with experience in both camps. The greatest differences between the film workflows and graphic arts workflows fall in areas of white point assumptions and total system dynamic range. Another area that is also common is the need to maintain a subjective "look and feel" across multiple media.
The largest population of users of serious ICC technology are "by the numbers" guys. A great deal of effort is spent making output devices behave in a more or less standard manner with definitions of primary, secondary colors and tone response curves. The device links that Graeme refers to are tables that resolve predicted color differences between device and an assumed standard input in device coordinates. This is exactly the same as the technology described by the troll you spoke with.
The motion picture industry tends to be driven by perceptual intent. Certainly, in video situations, real time editing is done and the ASC (cinematographers) have a standard set of controls that are simple, but defined to convey the intent of the director or cinematographer to the next stage of production.
From a display calibration standpoint, there are two very important factors in video and mp applications that generally don't exist in ICC applications. First there are two distinct standards that are generally used in these applications: Rec. 709 and DCI-P3. These are two prevailing output assumptions. Displays used in video applications often use a Serial Data Interface (SDI) which guarantees that each display receives exactly the same data. This means that these displays must contain hardware calibration capabilities inside the display. This same requirement is developing for video walls and ON-set applications where multiple folks have to see the same image, but are not physical proximity. This means that the calibration hardware transform is built into the display. Many high end displays have this capability and you will soon see it in low cost LCD displays as well. These transform engines in the display generally are 14+bits and use either table lut or lut-matrix architectures to transform the native display primaries to the required working space requirements. These are exactly analogous to the Device Link that Graeme refers to.
Modern display calibration is far more common between video, motion picture and graphic arts than ever before. The physical difference between displays is driven more by the physics of the display than any color management style. The motion picture applications push towards very wide dynamic range and the graphic arts applications push towards lower contrast ratios. This difference is aimed primarily at the differences in destination media.
There certainly is a lot of hubris in the Video and Motion Picture camp, but in the end, the approaches to get to the final result are far more common than different.
Regards, Tom Lianza
Steve Kale wrote: Hi,
My take on this is that the video guys 'hard-tune' devices to a specific primary/gamma set (to use terms loosely) whereas the ICC approach has a much higher degree of flexibility in that each device is profiled and each piece of media carries with it a profile that, in effect, contains expectations as to how that content should look.
Not really. The ICC approach breaks the task down into components, so that there is the flexibility to mix and match. Hence devices are characterised, and then the CMM links profiles to create an emulation (or adaptation) of one colorspace to another. The typical Video approach jumps directly to a single target, the desired response of the display, and then used more add-hock methods to calibrate the display so that it emulates the desired response. When the display and the target response are both additive, it's not too hard to break this down into hitting the desired primary chromaticities and per channel responses. If the device (say) is not close to additive, then this approach doesn't work too well, nor does it provide any mechanism to control out of gamut mapping (typically out of gamut colors will be clipped in some crude fashion, although the aim will be to avoid this by ensuring that the display has a greater gamut than the target response). In this case it's necessary to use an ICC like approach of calibrating the display to get it close and have sufficient gamut, characterise it and then link the desired response with the display characterisation to compute a device link (3D Lut) that implements the emulation. One of the advantages having a single specific target in mind has is that some fine tuning can be applied over the whole system, whereas the normal ICC approach puts up with any slight errors introduced in the separate characterisation of the input and output devices and the linking process.
display. It would seem much more sensible to me for the video industry to move to an ICC-profile type setup where recording devices are profiled (perhaps less important because of post processing), displays are profiled and each video (however processed for effect) comes with a profile that aids a CMM to deliver the correct look on any display.
Ideally, but I can very well understand this approach not being practical in the real world at present. The ICC approach has struggled (and still struggles) to work in a practical sense of profiles and systems being available, whereas fixed standards such as sRGB and the typical Video industry approach around Rec. 709 etc. are relatively simple and practical. They do straight jacket technical development though, resulting in great struggles when things like wider gamut displays or higher dynamic range are introduced. The utopia of Video being tagged with a profile would rely on most displays having 3D cLUT mapping hardware (or the equivalent), some reasonable "smart" CMM like linking and gamut mapping in every display, and a reasonably accurate profile of the display itself. This is a perfectly doable scheme that would work, the greatest obstacle being getting industry agreement, standardisation and a feely usable reference implementation. I guess it could happen in the future, but I suspect many people in the Video industry don't know enough about color to know that they don't know everything about color, and the current arrangements work "well enough".
Is this a fair view of the world? (says someone who has just profiled his computer monitor with i1Profiler while reading Poynton's "Digital Video and HD" and about to recalibrate my plasma display with HCFR)
It is, but the world is a messy place. Graeme Gill.
One of the advantages having a single specific target in mind has is that
some fine tuning can be applied over the whole system, whereas the normal ICC approach puts up with any slight errors introduced in the separate characterisation of the input and output devices and the linking process.
display. It would seem much more sensible to me for the video industry to move to an ICC-profile type setup where recording devices are profiled (perhaps less important because of post processing), displays are profiled and each video (however processed for effect) comes with a profile that aids a CMM to deliver the correct look on any display.
Ideally, but I can very well understand this approach not being practical in the real world at present. The ICC approach has struggled (and still struggles) to work in a practical sense of profiles and systems being available, whereas fixed standards such as sRGB and the typical Video industry approach around Rec. 709 etc. are relatively simple and practical. They do straight jacket technical development though, resulting in great struggles when things like wider gamut displays or higher dynamic range are introduced.
Very well said, I always think the failure of ICC workflow implementation in the industry in general comes from the lack of solid guidance/measure to match its confusion created by its flexibility and openness.
participants (3)
-
Graeme Gill
-
Nipat Paiboonponpong
-
Steve Kale