Approximating color temperature with smart phones and camera

46 posts / 0 new
Last post

Pages

maukka
maukka's picture
Offline
Last seen: 19 hours 33 min ago
Joined: 12/31/2015 - 04:15
Posts: 1817
Location: Finland
Approximating color temperature with smart phones and camera

Color temperature (cool, neutral, warm, measured in Kelvins) among tint (magenta, green) and color rendering (CRI) seems to be one of the most important factors when choosing a flashlight. Since it is usually difficult to measure the absolute color temperature without professional tools, I decided to test a couple different ways of approximating it with cheap methods. As a reference I used a X-Rite i1Pro spectrophotometer, to which all of the contenders were compared.

For meaningful results, I chose several different flashlights with differing color temperature and CRI. Although in the end CRI didn’t seem to correlate with the accuracy of any of the meters.

With all the methods except the i1Pro and i1Display Pro, which measure the ambient light directly from the source, I used a grey card meant for correcting white balance in photography. I pointed the flashlight on the grey surface and took the photo of the middle of the beam. I tried to maximize the hotspot so that it covered as much of the frame as possible.

The methods

Reference: X-rite i1Pro spectrophotometer. You need something like this to measure CRI and spectral distribution. (~1000 €).
https://www.xrite.com/product_overview.aspx?ID=1461

Colorimeter: X-rite i1Display Pro. A sophisticated lux meter with three sensors and colored filters. Can be used to calibrate displays. (~230 €)
http://www.xrite.com/i1display-pro

With the X-rite devices, I used Argyll CMS’s spotread.exe to measure color temperature (CCT).

For Android (Motorola Moto E 2014), I chose the White Balance Color Temp Meter downloadable from the Play Store (0,79 €).
https://play.google.com/store/apps/details?id=cassiopeia.camera&hl=fi

For iOS (iPhone 6), I chose White Balance Meter (2,99 €)
https://itunes.apple.com/fi/app/white-balance-meter/id834425480?mt=8

As a camera I used a Panasonic DMC-GM1 mirrorless DSLR and Adobe Photoshop Camera Raw converter. After taking the photo, I opened it up in Camera Raw and adjusted the white balance with the sample tool from the grey card. The software then shows the approximated color temperature.

Results
First table has measurements, the second has error in percents. Green background indicates an error under 5 %, red larger than 10 %.

Conclusion

Only the X-rite i1Display Pro gives consistent results, albeit a bit on the high side. It is well worth the price if you also want to calibrate your computer monitors, televisions and projectors.

The Android app and camera methods are not far from the i1Display Pro in average accuracy, but their results are wildly more unpredictable. The app is much faster to use, since copying images to computer and a RAW file processing software is not needed. Both are somewhat inconsistent, but work better than estimation by eye.

The iOS app I tried is absolute rubbish. None of the readings are accurate.

If you have suggestions for more apps, I’ll gladly test them as long as they are only a buck or two.

Edited by: maukka on 02/22/2016 - 03:40
emarkd
emarkd's picture
Offline
Last seen: 6 hours 12 min ago
Joined: 04/14/2015 - 22:04
Posts: 1567
Location: Georgia, USA

This is awesome, thank you for doing this! I can’t afford a thousand bucks for a sophisticated light meter, although I wish I could, but I can happily spend a few dollars to try some Android apps. Trouble is I have no way of knowing how accurate the are. So thanks for doing the comparison!

sp5it
sp5it's picture
Offline
Last seen: 7 hours 50 min ago
Joined: 12/25/2012 - 07:51
Posts: 773
Location: Poland

I have I1 Smile
Great info.
Subscribed.
Mike

Think of how stupid the average person is, and realize half of them are stupider than that. George Carlin

 Anyone offended by my signature please fill Complaint Form. Thank you.

sixty545
Offline
Last seen: 4 months 1 day ago
Joined: 10/25/2010 - 14:15
Posts: 702
Location: Denmark (GMT + 1)

Very interesting. I Want to buy that Android application for fast checks.
I think that your percentage scheme should be calculated using Mired, but the difference will most likely be small.

maukka
maukka's picture
Offline
Last seen: 19 hours 33 min ago
Joined: 12/31/2015 - 04:15
Posts: 1817
Location: Finland
sixty545 wrote:
I think that your percentage scheme should be calculated using Mired, but the difference will most likely be small.

Here you go. A mired shift of under 10 MK^-1 in green, over 20 MK^-1 in red.

djozz
djozz's picture
Offline
Last seen: 49 min 19 sec ago
Joined: 09/07/2012 - 17:04
Posts: 14963
Location: Amsterdam

Nice measurements, I love reading this kind of stuff! There’s many things to find out once you have a good reference for tint and CRI, and I really appreciate that you are doing that and take the trouble to post about it Smile

I assume btw that the quality of the colour temperature reading of a phone is not just app-dependent but also may vary with the measuring hardware on your phone, the same app may give results of different quality on different smartphones.

A bit related: there have been a few attempts on BLF to use a smartphone as luxmeter, this was my latest ‘pilot study’ last year: http://budgetlightforum.com/node/39477

Keep up the good work!

maukka
maukka's picture
Offline
Last seen: 19 hours 33 min ago
Joined: 12/31/2015 - 04:15
Posts: 1817
Location: Finland

Yes, I should have mentioned that the phone used was Motorola Moto E (2014) for the Android app and iPhone 6 for the iOS.

sixty545
Offline
Last seen: 4 months 1 day ago
Joined: 10/25/2010 - 14:15
Posts: 702
Location: Denmark (GMT + 1)

Thanks for the calculation work in Mired! That gives me a better “feeling” about the results.
For people who want to know what Mired is, read about it on wikipedia.org

joelbnyc
joelbnyc's picture
Offline
Last seen: 8 months 3 days ago
Joined: 04/19/2013 - 21:07
Posts: 43
maukka wrote:
sixty545 wrote:
I think that your percentage scheme should be calculated using Mired, but the difference will most likely be small.

Here you go. A mired shift of under 10 MK^-1 in green, over 20 MK^-1 in red.

!http://i.imgur.com/2SCuL7y.png!

I know this is old, but I’ll add that the $170USD XRite Colormunki is supposed to be about as accurate a display calibrator as the Pro. We bought it few years ago for SO’s photography business, as any digital graphic artist is working blind if their displays aren’t calibrated.

I leave my own displays uncalibrated, though. For general use, the factory calibration is often more pleasing than accuracy. Just don’t imagine your images will look the same on all devices, or print accurately. Or render beamshot tints especially accurately. And I don’t even bother to print images I’ve taken, because I know only images like hers, shot in Raw and edited on a calibrated monitor, will print similarly to how they appear on the screen.

contactcr
Offline
Last seen: 5 hours 7 min ago
Joined: 05/19/2017 - 18:52
Posts: 1706
Location: US

So I think the ColorMunki Photo + Argyll will give full spectral data and would be the least costly solution.

So CCT, CRI, DUV, etc right since it’s a spectrophotometer and not simply a colorimeter. Is this right?

maukka
maukka's picture
Offline
Last seen: 19 hours 33 min ago
Joined: 12/31/2015 - 04:15
Posts: 1817
Location: Finland

contactcr wrote:
So I think the ColorMunki Photo + Argyll will give full spectral data and would be the least costly solution.

So CCT, CRI, DUV, etc right since it’s a spectrophotometer and not simply a colorimeter. Is this right?

Yes, you’re correct. Also goes for the new i1Studio.

dotBLF
Offline
Last seen: 6 hours 57 min ago
Joined: 04/19/2016 - 19:00
Posts: 55
Location: MA, USA

Wicked!

I’ve got the i1 so this is definitely something I’m going to be checking out, great stuff as always maukka!

contactcr
Offline
Last seen: 5 hours 7 min ago
Joined: 05/19/2017 - 18:52
Posts: 1706
Location: US

maukka wrote:
contactcr wrote:
So I think the ColorMunki Photo + Argyll will give full spectral data and would be the least costly solution.

So CCT, CRI, DUV, etc right since it’s a spectrophotometer and not simply a colorimeter. Is this right?

Yes, you’re correct. Also goes for the new i1Studio.

I ordered a used one from eBay for $250, will play around with it and report back!

contactcr
Offline
Last seen: 5 hours 7 min ago
Joined: 05/19/2017 - 18:52
Posts: 1706
Location: US

ColorMunki Photo arrived.

I got the Argyll drivers installed and began with “spotread -a” after reading only a fraction of the documentation.

I hope maukka can provide a crash course tutorial for me/us. Some initial questions:

  • Should I be concerned about distance from my light to meter?
  • How much do I care about ambient light? My room is fairly dark and I put the device under my desk w/ monitor on and shine light from desktop height?
  • What spotread command works best for our flashlights?
  • What is it doing during calibration? Do I need to have it set up any special way?
  • It said (Caution) in the output on some lights/modes but still gave me somewhat reasonable output. Should I be concerned?
  • Continued from above, will I run into a lux ceiling and how can I adjust to measure higher output if so?

Any other best practices would be appreciated. For a ~$250 outlay if I can get my dual monitor setup calibrated at work (same model monitor but different panels and ~5 years apart) and still use it to screw around with LEDs seems like a pretty “fun” investment.





219C 4000K 90CRI w/ smooth narrow carclo @ high-ish amps, measured from about 2.5 feet away:

Quote:
Unable to apply FWA compensation because it wasn’t set up Result is XYZ: 5474.103559 5464.804218 3681.450380, D50 Lab: 424.199480 24.285244 49.547610 Ambient = 5464.8 Lux, CCT = 4148K (Duv 0.0004) Suggested EV @ ISO100 for 5464.8 Lux incident light = 11.1 Closest Planckian temperature = 4136K (DE2K 0.7) Closest Daylight temperature = 4209K (DE2K -3.8) Color Rendering Index (Ra) = 91.8 [ R9 = 65.8 ] R1 = 91.4 R2 = 94.4 R3 = 95.5 R4 = 90.7 R5 = 90.3 R6 = 90.4 R7 = 94.7 R8 = 86.7 R9 = 65.8 R10 = 85.5 R11 = 89.3 R12 = 71.5 R13 = 92.0 R14 = 97.6 Television Lighting Consistency Index 2012 (Qa) = 92.3

Luxeon MZ 3000K + 5700K in Cute4 medium optic:

Medium output:

Quote:
Unable to apply FWA compensation because it wasn’t set up Result is XYZ: 3855.196692 3885.619571 2532.561524, D50 Lab: 376.896429 16.218168 51.156372 Ambient = 3885.6 Lux, CCT = 4154K (Duv 0.0022) Suggested EV @ ISO100 for 3885.6 Lux incident light = 10.6 Closest Planckian temperature = 4092K (DE2K 3.4) Closest Daylight temperature = 4170K (DE2K -1.0) Color Rendering Index (Ra) = 92.5 [ R9 = 79.6 ] R1 = 93.0 R2 = 92.8 R3 = 91.2 R4 = 93.5 R5 = 91.9 R6 = 88.9 R7 = 95.7 R8 = 93.1 R9 = 79.6 R10 = 82.3 R11 = 92.8 R12 = 72.7 R13 = 92.4 R14 = 95.1 Television Lighting Consistency Index 2012 (Qa) = 95.3

High ~12A output:

Quote:
Unable to apply FWA compensation because it wasn’t set up Result is XYZ: 16710.503957 16757.760492 10929.855626, D50 Lab: 623.533525 31.075066 83.035566 Ambient = 16757.8 Lux, CCT = 4117K (Duv 0.0015) Suggested EV @ ISO100 for 16757.8 Lux incident light = 12.7 Closest Planckian temperature = 4075K (DE2K 2.4) Closest Daylight temperature = 4150K (DE2K -2.1) Color Rendering Index (Ra) = 92.3 [ R9 = 77.2 ] R1 = 92.8 R2 = 92.8 R3 = 91.3 R4 = 93.2 R5 = 91.8 R6 = 89.0 R7 = 95.4 R8 = 92.1 R9 = 77.2 R10 = 82.2 R11 = 92.6 R12 = 73.7 R13 = 92.3 R14 = 95.2 Television Lighting Consistency Index 2012 (Qa) = 94.4
maukka
maukka's picture
Offline
Last seen: 19 hours 33 min ago
Joined: 12/31/2015 - 04:15
Posts: 1817
Location: Finland
  • Should I be concerned about distance from my light to meter?

Because the meter also works as a lux meter, if you want to measure cd you need to know the distance to convert from lux (cd = lux*(distance in meter)^2, throw in meters = square root(cd/0.25))

  • How much do I care about ambient light? My room is fairly dark and I put the device under my desk w/ monitor on and shine light from desktop height?

Ambient light is usually so minuscule in intensity in relation to the light you’re measuring that it usually has no effect on the results.

  • What spotread command works best for our flashlights?

Looking at the printout it seems you got that already figured out, I’d just add -H for hires mode and -x for xy coordinates in the CIE space Thumbs Up

  • What is it doing during calibration? Do I need to have it set up any special way?

It calibrates for dark reading. No need to do anything else than put the device in calibration mode (rotate the side wheel to the correct orientation)

  • It said (Caution) in the output on some lights/modes but still gave me somewhat reasonable output. Should I be concerned?

The measurement might have been clipped at some wavelengths which could skew the results.

  • Continued from above, will I run into a lux ceiling and how can I adjust to measure higher output if so?

Increase distance.

  • Any other best practices would be appreciated. For a ~$250 outlay if I can get my dual monitor setup calibrated at work (same model monitor but different panels and ~5 years apart) and still use it to screw around with LEDs seems like a pretty “fun” investment.

It’s as simple as that. And fun!

contactcr
Offline
Last seen: 5 hours 7 min ago
Joined: 05/19/2017 - 18:52
Posts: 1706
Location: US

Thanks for all the responses. I will try to refine my testing a bit to make sure everything is consistent. I did test the calibration at work and it got my 2 displays calibrated closer to each other in 10 minutes than I could achieve with publicly uploaded ICC profiles and manual adjustments after hours. Looks to be a very versatile device.

I have a USB on the go cable I will try with an android tablet and ArgyllPro ColorMeter DEMO version as well as BabelColor evaluation copy. Are there any open source or free/cheap things to try? Argyll’s command line tools seem pretty full featured but they don’t show the same eye candy of your BabelColor screen captures!

maukka
maukka's picture
Offline
Last seen: 19 hours 33 min ago
Joined: 12/31/2015 - 04:15
Posts: 1817
Location: Finland

HCFR is free and it can plot the xy coordinates and show a colorful spectrum image.

contactcr
Offline
Last seen: 5 hours 7 min ago
Joined: 05/19/2017 - 18:52
Posts: 1706
Location: US

I thought i read somewhere you could make your own spectrum image with the XY data, can you elaborate? Maybe i can just make my own program to read .sp output and plot exactly what i need

maukka
maukka's picture
Offline
Last seen: 19 hours 33 min ago
Joined: 12/31/2015 - 04:15
Posts: 1817
Location: Finland

You can graph the spectrum in Excel for example, but not using the CIExy data. That’s just the tint coordinate. Spotread -s gives you the necessary wavelength data for that. Or plots the spectrum with -S.

bmengineer
bmengineer's picture
Offline
Last seen: 2 hours 40 min ago
Joined: 01/26/2018 - 11:38
Posts: 530
Location: Ontario, Canada

Is there any way of knowing how accurate the Aperture Meter Android app is?

Find all my reviews of flashlights and more gear at www.bmengineer.com

maukka
maukka's picture
Offline
Last seen: 19 hours 33 min ago
Joined: 12/31/2015 - 04:15
Posts: 1817
Location: Finland

It has a calibration feature for CCT and lux so it could work. I’ll have to check it out.

edit: did some tests, on my phone (the ancient Motorola Moto E) the CCT reading doesn’t work since the phone doesn’t have an RGB front light sensor and its lux readings are all over the place and wildly vary with different CCT and CRI sources. So, may or may not work depending on the device.

contactcr
Offline
Last seen: 5 hours 7 min ago
Joined: 05/19/2017 - 18:52
Posts: 1706
Location: US

I see the spectral data in the file but how do you go from that to this? or maybe I just need this to be written to the file so I can parse it?

Yxy: 7578.662723 0.364250 0.361470

Ambient = 7578.7 Lux, CCT = 4375K (Duv -0.0022)

Color Rendering Index (Ra) = 89.9 [ R9 = 77.8 ]
R1 = 92.2 R2 = 90.6 R3 = 87.2 R4 = 90.1 R5 = 91.2 R6 = 86.0 R7 = 91.1 R8 = 91.0 R9 = 77.8 R10 = 77.0 R11 = 90.9 R12 = 73.9 R13 = 91.1 R14 = 92.8

Trying to do something like you do here:

maukka
maukka's picture
Offline
Last seen: 19 hours 33 min ago
Joined: 12/31/2015 - 04:15
Posts: 1817
Location: Finland

That’s just a spreadsheet plotting those xy coordinates.

edit: you could also check out Osram Color Calculator.

contactcr
Offline
Last seen: 5 hours 7 min ago
Joined: 05/19/2017 - 18:52
Posts: 1706
Location: US

but i only see those coordinates in the spotread output not in the spectrum file it saves

maukka
maukka's picture
Offline
Last seen: 19 hours 33 min ago
Joined: 12/31/2015 - 04:15
Posts: 1817
Location: Finland

I usually just take a single measurement and log the data by piping the output to a file, because spotread doesn’t save all the data in the log. Or just copy and paste the text from the command line window. After first calibrating, use whatever parameters and -N -O > log.txt

bmengineer
bmengineer's picture
Offline
Last seen: 2 hours 40 min ago
Joined: 01/26/2018 - 11:38
Posts: 530
Location: Ontario, Canada

Is it likely that the accuracy is device dependent?

Find all my reviews of flashlights and more gear at www.bmengineer.com

contactcr
Offline
Last seen: 5 hours 7 min ago
Joined: 05/19/2017 - 18:52
Posts: 1706
Location: US

This free program Osram ColorCalculator is super cool. I re-arranged the spectrum data into a CSV and imported it:

Here is a SST-20 3000K on high with a frosted optic

CIE plot:

R-value plot:

Some report like babel shows:

The actual program window:

DavidEF
DavidEF's picture
Offline
Last seen: 6 hours 14 sec ago
Joined: 06/05/2014 - 06:00
Posts: 7012
Location: Salisbury, North Carolina, USA

Can somebody tell me how to calibrate the colors on a projector connected to one of those stupid fruit-branded all-in-one desktop computers? At our church, the media computer is a Mac of some sort, a few years old. The projector is connected by HDMI over Thunderbolt, IIRC. The colors are horrible! Blues are over-saturated while reds are badly washed-out. I tried the on-board “monitor calibration” …thing on the Mac, but it only had brightness, contrast, maybe gamma, but no per-color-channel fixing. The projector itself has per-color-channel controls, but they don’t help, so the problem seems to be in the signal from the Mac! The LCD monitors connected to the Mac show colors “fine” though. Just the projector doesn’t. So, I don’t know what the deal is actually. Help?

The Cycle of Goodness: “No one prospers without rendering benefit to others”
- The YKK Philosophy

maukka
maukka's picture
Offline
Last seen: 19 hours 33 min ago
Joined: 12/31/2015 - 04:15
Posts: 1817
Location: Finland
contactcr wrote:
This free program Osram ColorCalculator is super cool. I re-arranged the spectrum data into a CSV and imported it:

I agree, it’s a great software for the price Smile

It seems their newest version also supports the TM-30-18 report format, which looks very professional. The latest CT&A also does that.

Barkuti
Barkuti's picture
Offline
Last seen: 1 hour 45 min ago
Joined: 02/19/2014 - 14:46
Posts: 3587
Location: Alhama de Murcia, Spain

contactcr wrote:

Luxeon MZ 3000K + 5700K in Cute4 medium optic:

Medium output:

Quote:
Unable to apply FWA compensation because it wasn't set up

Result is XYZ: 3855.196692 3885.619571 2532.561524, D50 Lab: 376.896429 16.218168 51.156372
Ambient = 3885.6 Lux, *CCT = 4154K (Duv 0.0022)*
Suggested EV @ ISO100 for 3885.6 Lux incident light = 10.6
Closest Planckian temperature = 4092K (DE2K 3.4)
Closest Daylight temperature = 4170K (DE2K -1.0)
*Color Rendering Index (Ra) = 92.5 [ R9 = 79.6 ]*
R1 = 93.0 R2 = 92.8 R3 = 91.2 R4 = 93.5 R5 = 91.9 R6 = 88.9 R7 = 95.7
R8 = 93.1 R9 = 79.6 R10 = 82.3 R11 = 92.8 R12 = 72.7 R13 = 92.4 R14 = 95.1
Television Lighting Consistency Index 2012 (Qa) = 95.3

High ~12A output:

Quote:
Unable to apply FWA compensation because it wasn't set up

Result is XYZ: 16710.503957 16757.760492 10929.855626, D50 Lab: 623.533525 31.075066 83.035566
Ambient = 16757.8 Lux, *CCT = 4117K (Duv 0.0015)*
Suggested EV @ ISO100 for 16757.8 Lux incident light = 12.7
Closest Planckian temperature = 4075K (DE2K 2.4)
Closest Daylight temperature = 4150K (DE2K -2.1)
*Color Rendering Index (Ra) = 92.3 [ R9 = 77.2 ]*
R1 = 92.8 R2 = 92.8 R3 = 91.3 R4 = 93.2 R5 = 91.8 R6 = 89.0 R7 = 95.4
R8 = 92.1 R9 = 77.2 R10 = 82.2 R11 = 92.6 R12 = 73.7 R13 = 92.3 R14 = 95.2
Television Lighting Consistency Index 2012 (Qa) = 94.4

Does this mean high colour accuracy light source tint mixing isn't really detrimental to their overall CRI, TLCI and such indexes? 

 

DavidEF wrote:
Can somebody tell me how to calibrate the colors on a projector connected to one of those stupid fruit-branded all-in-one desktop computers? … So, I don't know what the deal is actually. Help?

Monitor test patterns may be of help, try here for example: The Lagom LCD monitor test pages

 

contactcr
Offline
Last seen: 5 hours 7 min ago
Joined: 05/19/2017 - 18:52
Posts: 1706
Location: US

Speaking of tint mixing you can enter multiple measurements and the program will plot an estimate for that too.

Unfortunately CT&A doesn’t appear to support the ColorMunki Photo, or at least i couldnt make it work

Pages