Hughes: Introducing the ColorHug open source colorimeter [LWN.net]

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 14, 2011 15:23 UTC (Mon) by epa (subscriber, #39769) [Link] (11 responses)

Great news. I've been meaning to calibrate a few monitors but I didn't want to shell out for a proprietary colorimeter with its restrictive driver software. Reading the description of the device it's simpler than I had expected. I suppose the secret is to make sure you have exactly the right shade of red, blue and green in the colour filters and that you know exactly what proportion of incoming light they let through.

Hopefully, too, there can be a database of colour profiles for common monitors which can be included with Linux distributions. Yes, every individual monitor is different and should ideally be calibrated individually; but still a general profile for all monitors of a certain type is better than no profile at all.

Perhaps the LiveCD software could ask if you want to submit your profile to a central database.

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 14, 2011 16:18 UTC (Mon) by slashdot (guest, #22014) [Link] (10 responses)

Isn't it much better to simply make monitors that correctly output the sRGB colorspace instead of being broken and requiring calibration?

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 14, 2011 16:32 UTC (Mon) by zlynx (guest, #2285) [Link]

Many of the more expensive monitors are calibrated after production and so they are "correct" as far as that goes.

I believe the problem comes in because the monitor also reflects some of the light in the room back at the viewer. This changes its color profile.

So to get the monitor perfectly calibrated you need to calibrate with the room lighting set up just as you would be using it.

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 14, 2011 16:45 UTC (Mon) by mattdm (subscriber, #18) [Link]

It seems unreasonably idealistic to hope that all displays would ship calibrated to any standard. However, sRGB in particular would be a very limiting choice since it is a rather restricted space.

It's okay as a fallback "safety", but by design it's a lowest-common-factor. Restricting all hardware to that would keep us from taking advantage of better hardware and getting the most out of new display technology.

By working with a color-calibrated device in the first place and using a color-managed workflow, you can take advantage of better hardware and be reasonably certain that the images you create will render decently on a "non-broken" default-sRGB device.

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 14, 2011 16:58 UTC (Mon) by renox (guest, #23785) [Link] (6 responses)

> Isn't it much better to simply make monitors that correctly output the sRGB colorspace instead of being broken and requiring calibration?

Uh? You know nothing about making monitors, right?
Even if monitors were 100% correctly calibrated initially, the color balance of monitors can change with time..

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 14, 2011 18:08 UTC (Mon) by drag (guest, #31333) [Link] (5 responses)

That's one thing: Monitors can change colors for quite a few reasons. So they will periodic recalibration.

=--=-=-=-=-

Plus there is no such thing as 'correct sRGB colorspace'. Colors change constantly. To think there is one set of 'correct colors' is just mistaken.

Print out a image on a piece of paper. Look at it under a incandescent light, now look at it under a floresent light, take it out in the morning sun and look at it, take it out into the afternoon sun and look at it, take it out in the evening sun and look at it. Look at with a gray back drop, now look at it with a blue back drop, etc etc.

Each time and each condition the paper will have radically different colors. Colors of everything change constantly because the colors are mental construct that are dependent on light striking a surface and shining through it.

As the light changes and our perception changes then so will the colors. Even with a monitor with a set light source behind it will have it's colors affected by the environment.

The point of color calibration is not to 'Find the correct colors' (although obviously you want your monitor to be correctly configured as much as possible). It's to try to color match a process, a work flow. So when you are working with other people and other media then you use your color calibration and color profiles to color match.

Example: if you are working with a video that is going to display output on a movie screen you can get very close to how it will appear on your monitor. Also you can make sure that when working with a dozen different people on different computers with different displays that the color red for you will be the same as them. Same thing with different inputs from different sources. All that stuff needs to be color matched.

Example 2: say you are working with a industrial printer. A industrial printer will use different sets of inks, each with different color capabilities based on how fast you want to print, what material you are printing on, and how much money you want to spend on ink. Each set of ink will have different color profiles and different colors that it's capable of representing. What makes it more difficult is that printing out test prints is prohibitively expensive sometimes. So you only have one shot to get it right. Color matching your monitor with your printer and also having a desktop layout program that will restrict the color gamuts to the correct ones is very important.

For all of this you will to be able to color calibrate all your outputs and inputs. Scanners, printers, cameras, your monitor.. each will have different capabilities and different color profiles.

it's a difficult problem that nobody has solved yet. But things like this 'colorhug' will help mitigate a lot of the issues artists have to deal with in professional environments using Linux.

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 14, 2011 19:01 UTC (Mon) by slashdot (guest, #22014) [Link] (4 responses)

> Plus there is no such thing as 'correct sRGB colorspace'

How so?

As far as I can tell, you can get CIE XYZ values by integrating a measured light distribution (after a constant linear transformation) weighted with the functions at http://en.wikipedia.org/wiki/File:CIE_1931_XYZ_Color_Matc..., convert to linear RGB via the matrix multiplication at http://upload.wikimedia.org/wikipedia/en/math/b/d/2/bd2b2... and finally each component to sRGB with the formula at http://upload.wikimedia.org/wikipedia/en/math/c/d/e/cde58...

If there is a point perpendicular to the center of the screen and an input-independent linear transformation of the light distribution measured from that point with no other light sources, such that the result of the above process rounded to an integer is the sRGB value provided as input, for any input and any pixel, the monitor works correctly, otherwise it doesn't.

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 14, 2011 21:30 UTC (Mon) by dgm (subscriber, #49227) [Link]

He was not talking about your screen working "correctly", but about getting to see in it the same colors you will get from another monitor, or your printer.

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 14, 2011 22:39 UTC (Mon) by Wol (subscriber, #4433) [Link] (2 responses)

How So?

Because YOUR EYES don't perceive colour consistently! So if you want a consistent experience, the monitor needs to change.

Try this experiment. Place a yellow banana on a piece of yellow paper. They're both yellow, aren't they ... ?

Now illuminate it under red light. The paper is now orange, but the banana is still yellow!

Colour perception is very much an ART, not a science (well, it is science, but it's random and statistical, not exact).

Cheers,
Wol

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 15, 2011 0:18 UTC (Tue) by nix (subscriber, #2304) [Link] (1 responses)

This is, of course, an example of *consistent* colour perception. The frequency mix of the light reflected from the banana has changed, but the *colour* is the same, because 'colour' is a thing entirely situated within the brain, and is corrected to a large degree for lighting conditions, shading, surface texture and much else. (Nothing we can build can do such a good job, let alone do it in realtime with noisy, constantly jittering input from input sensors laid out partly at random.)

stochastic sensor placement

Posted Nov 15, 2011 14:30 UTC (Tue) by tialaramex (subscriber, #21167) [Link]

The "input sensors laid out partly at random" is actually a clever trick, which we do in fact use in things we build although not cameras so far as I know. The sensors are in fact arrayed stochastically, which inherently avoids artefacts like aliasing.

Suppose you're looking at, for example a distant checkerboard pattern. If you use a perfect rectangular grid of sensors at certain distances, you will "see" patterns that don't really exist because of the regularity of your sampling. If you instead "randomly" place the sensors according to a rule that limits their proximity you completely avoid this problem.

Pixar's Photorealistic Renderman (and presumably many other modern 3D rendering systems) likewise uses multiple stochastically chosen sample origin points for each pixel, with the same result - no aliasing.

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 16, 2011 1:17 UTC (Wed) by ringerc (subscriber, #3071) [Link]

sRGB is horrible. It was designed as the least-common-denominator colour space for *CRT* monitors. Not only is it a narrow crappy gamut, but it's inappropriate for LCD displays.

In any case, unless you bought a rather high end display designed for colour work, your monitor won't be factory-calibrated for sRGB, it'll have a canned set of pixel gains pre-configured for the entire production run. At best these will be inaccurate; most of the time they're totally bogus.

sRGB is not the answer unless you do no colour-sensitive work whatsoever. Even then, it's ugly.

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 14, 2011 20:21 UTC (Mon) by job (guest, #670) [Link] (1 responses)

Is somebody knowledgeable enough to determine if this is a viable construction? Will it be competitive with other devices?

I'd be more than happy to buy a device to support open source hardware, but it must work in practice too. I have just no idea how these things work. The device itself looks disturbingly simple. Is this all these things are? Proprietary calibration devices are several times more expensive.

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 15, 2011 17:23 UTC (Tue) by Cyberax (✭ supporter ✭, #52523) [Link]

Yes, it's a viable construction and it can be competitive with other devices. Although I doubt it's going to be significantly cheaper.

Now if only we could get reflected-light spectrometers to make color profiles of printers and paper... The cheapest ones cost about $1000 right now and are a total crap.

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 14, 2011 23:26 UTC (Mon) by tetromino (guest, #33846) [Link] (4 responses)

The pricing makes it non-viable, at least for US residents.

A ColorHug, even with the developer discount, costs $81 with the cheapest shipping option to the US. Pantone Huey (which as Hughes himself states has more features) can be had for as little as $65 with shipping. And the Huey has Windows drivers too, which is important if I want to let my friends or family borrow the colorimeter.

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 15, 2011 0:24 UTC (Tue) by pabs (subscriber, #43278) [Link]

Since it is open hardware you can probably find a way to make it yourself in the USA or find someone willing to do that for you. There is no reason Hughes has to be the only person building the things!

I guess he intends to add more features too, you might want to contact him to find out if Windows drivers on the to-do list, if I were him I personally would not bother with them.

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 15, 2011 11:49 UTC (Tue) by hpro (subscriber, #74751) [Link]

Well, according to quite a few of the comments on pantones page about the Huey it does not seem to get new updates anymore, making it useless for new-enough systems..

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 15, 2011 20:13 UTC (Tue) by mattdm (subscriber, #18) [Link]

Richard has said that the current pricing is for prototyping a small run. A much larger run could in theory be much, much cheaper.

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 16, 2011 1:04 UTC (Wed) by ringerc (subscriber, #3071) [Link]

A $16 difference doesn't seem non-viable to me, especially when comparing to currently in-production colorimeters like the i1Display2 and the (ick) Spyder rather than the obsolete/unsupported Huey.

Sadly for my work I really need a spectro, so a plain colorimeter won't cut it. I may pick one up to get involved with testing and to encourage work on a spectro, though.

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 15, 2011 17:12 UTC (Tue) by jcm (subscriber, #18262) [Link] (1 responses)

This is plain awesome. No other way to describe it.

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 16, 2011 13:28 UTC (Wed) by drag (guest, #31333) [Link]

yes. Yes it is awesome.

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 16, 2011 1:13 UTC (Wed) by ringerc (subscriber, #3071) [Link] (3 responses)

Wondering why you need one of these?

Unless you do accurate colour work, you don't. If you do accurate colour work, though, it's vital.

A quick and dirty FAQ about colorimeters and calibration in general (rather than this one in particular):

Q: "My monitor already comes with a profile, so why should I create one?"

A: Actually, at best your monitor comes with a profile that was vaguely appropriate for the model when it was first produced. It doesn't come with a profile for your particular panel as each individual panel and backlight is different. If your monitor isn't brand new the profile is out of date because the panel's colour changes as it ages. In any case, unless you have a high end professional colour panel it's likely that the profile that came with it is completely bogus or is just a dummy sRGB profile that's effectively a no-op.

Q: "My monitor has an sRGB mode, so why can't I just use a regular sRGB profile and avoid having to calibrate?"

A: The sRGB mode in your monitor is incredibly unlikely to have been measured and calibrated during production. It'll be a generic set of settings for the entire production run. Even if yours is one of the high end panels where the sRGB mode was actually calibrated for each individual panel during production, it's out of date now as the monitor has aged. Worse, sRGB was designed for CRTs so it has a narrow gamut that's inappropriate for LCDs and restricts the colours that can be displayed.

Q: "Can I use this to calibrate printers and scanners too?"

A(1): For printers you need a spectrophotometer. They're expensive, like $lots, because they require precision calibration during production, fancy emissive light sources, and more. Try to find a ColorMunki or a 2/h i1Pro but if it's an older instrument be prepared to pay to have it recalibrated.

A(2): For scanners you can get a cheap but excellent quality IT8 target from Wolf Faust (http://targets.coloraid.de/), scan it, and calibrate using Argyll CMS, VueScan, or your vendor scanner software if it isn't totally crap. I cannot recommend Wolf's targets enough.

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 18, 2011 13:39 UTC (Fri) by dafid_b (guest, #67424) [Link] (2 responses)

I have an idea.. and would appreciate the constructive feedback you want to give.

First some background: I have a student artist partner, who was introduced to colour profiles for printers etc last year. I have read a very little in the popular press about colour profiles and a tiny amount in programming APIs about colour gamuts - so am not an expert in any way!

Between us we have three monitors, a scanner, a couple of digital cameras, a couple of light sources (sun-light photo lamps) and a photo-printer which we use with cheap paper, quality paper, cheap ink and photo ink.

We could get some IT8 targets, thanks for that link:), and could calculate profiles for the scanner and cameras as described.

However, could the ColorHug and spectrophotometer be replaced by a process like the following?

1.a) Scan the IT8 target image
1.b) Take a photo of the target image under photo-light-a with camera-a
1.c) Take a photo of the target image under photo-light-a with camera-b
1.d) Take a photo of the target image under photo-light-b with camera-a
1.e) Take a photo of the target image under photo-light-b with camera-b
1.f) Take a photo of the target image under sun-light(1?) with camera-a
1.g) Take a photo of the target image under sun-light with camera-b

[1? Is variation in sunlight due to weather, season, or pollution something that can be corrected for? For important events, today's sunlight could be used as a prediction for tomorrows sunlight:)]

Then process the results and make colour profiles for these combinations.

2) print the target image on quality paper

Then scan and photograph the output of the printer using the calibrated camera and scanner...

3.(a...g) calibrate all the input devices using the printed output

4) print the target image on cheap paper

Then scan and photograph the output of the printer using the calibrated camera and scanner...

5.(a...g) calibrate all the input devices using the printed output

(and etc for the cheap ink and other variations)

For the monitors...

1) Display the reference image.
2) Take long exposure (to average out any flickering) with each digital camera
3) Calculate profiles

The above sounds complicated, and laborious. However it provides colour profiles for most of the anticipated uses for the many devices - and the user can choose which combinations make sense for their use of the devices. As such it seems like a reasonable approach.
With the application of some automation it should not be too difficult to manage.

The 'A calibrates B which calibrates C' would introduce errors, but because of the large number of devices used, the mess of cross-measurements would provide opportunities for error minimisation or correction that would counter these introduced errors - another task for software.

Also, there are comments above about the 'colour profile' of a monitor changing with age, placement in a room and the lighting that surrounds it. I think the photo of the image can capture these impacts - whereas the ColorHug might not be able to?

It would be nice to have a midday profile, an afternoon profile and a night profile for the monitors and also sunlight, fluorescent and incandescent viewing profiles for the print-outs - although this is more about an integrated colour management facility in the desktop than about how to obtain a colour profile...

Thanks for reading so far, looking forward to any comments!

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 20, 2011 12:20 UTC (Sun) by ringerc (subscriber, #3071) [Link]

In theory you can use processes like that; the common application is to profile a printer using a scanner that's been profiled with an IT8 target. The scanner is used to scan the printed swatches.

There are several problems with this. One is multiplication of error - deltaE is greatly increased by "chaining" profile production like this, greatly reducing the quality of the resulting profile.

If the gamut of the intermediary device (in this example a scanner) doesn't wholly encompass the gamut of the device being profiled then you'll have no useful data for the areas where the intermediary can't capture the profiled device's output. For example, your profile might not think your printer can to bright orange because the scanner can't capture bright orange so the bright orange swatches in the printout looked reddy-brown instead and the profile generator considered them out of gamut.

Perhaps more importantly, the devices don't always use the same colour models - for example, an RGB scanner might be being used to profile a printer that uses CMYK inks. That'll make it really hard to get a good profile and will require different logic in the profiling software to make a profile that's actually useful with the output device.

Overall, it's an OK-ish way to make rough profiles of devices if you don't have any better options, but there are plenty of reasons it's a crap way to do profiling and not generally recommended.

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 24, 2011 13:13 UTC (Thu) by gwg (guest, #20811) [Link]

The main problem with attempting to use a scanner or camera is that
it isn't a colorimeter. Typically such devices have spectral
sensitivities that are noticeably different to the human eye,
hence their measurement of color doesn't match what we see very well.
A color measurement instrument on the other hand is designed to emulate
the spectral sensitivities of the human eye, and it's success at this
is a measure of such an instruments quality (and cost!).

You can certainly try this camera or scanner type of approach out
if you like (see <http://www.argyllcms.com/doc/Scenarios.html#PP4>),
but the chances are that you will be disappointed in the
result.

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 17, 2011 10:31 UTC (Thu) by pabs (subscriber, #43278) [Link] (1 responses)

Apparently there was another one of these before:

http://www.homecinema-fr.com/colorimetre/index_en.php

Hughes: Introducing the ColorHug open source colorimeter

Posted Nov 24, 2011 13:18 UTC (Thu) by gwg (guest, #20811) [Link]

Yes, and it didn't work that well as a monitor instrument. The
main issues with it were it's lack of infra-red filters,
the very wide acceptance angle (which typically causes LCD
displays to "look too light" to the instrument, due to LCD
viewing angle dependence), and the lack of calibration for
particular displays. The latter is needed if the instrument does
not have color filters that accurately duplicate the human
eyes spectral sensitivities. This is typically the case
with cheap RGB type sensors. This can be overcome to a large
degree if the instrument is calibrated specifically for
each type of display's colorant spectra, but setting up
a cheap, efficient and accurate means of doing this calibration
is a challenge.