True Colour Recognition

Anything not specifically related to NBTV, but at least of some technical nature that might be of interest to NBTV members. Items for sale and links to retailers do not belong here.

Moderators: Dave Moll, Steve Anderson

Postby Klaas Robers » Thu Apr 18, 2013 3:52 am

Gary, search on spectroscope or spectrometer. These instruments work on a prism and give you the spectrum of the visible part of the light.

In the 70's, when I worked at Philips Research, some kind of artist wokred there too in order to develop a paint mixing system, to make paints for backgrounds in CTV scenes where you can predict the RGB value of your mixed paint. At least this was for the Philips Plumbicon CTV cameras and in incandescent studio lighting.

He did many mixing experiments and then went to a spectrometer (of course we had that on the lab) and came back with the spectrum curves in order to see what he did.
User avatar
Klaas Robers
"Gomez!", "Oh Morticia."
 
Posts: 1656
Joined: Wed Jan 24, 2007 8:42 pm
Location: Valkenswaard, the Netherlands

Postby M3DVQ » Thu Apr 18, 2013 4:24 am

A spectrometer will tell you what wavelengths are being reflected, but measuring their intensity accurately is very difficult.

I would be very interested to know how those paint matching scanners work too. It's possible they use the same principle as the "toy" ones but with more accurate and calibrated optoelectonics.
M3DVQ
Just nod and pretend you understand me
 
Posts: 338
Joined: Sat Feb 10, 2007 10:52 am
Location: Lincolnshire

Postby gary » Thu Apr 18, 2013 9:57 am

Klaas Robers wrote:Gary, search on spectroscope or spectrometer. These instruments work on a prism and give you the spectrum of the visible part of the light.


That's an interesting thought Klaas, I suppose the problem might be that an accurate spectrometer would be as expensive as an industrial colour recognition system but it bears further investigation.

M3DVQ wrote:A spectrometer will tell you what wavelengths are being reflected, but measuring their intensity accurately is very difficult.


Oh, that would be a problem yes.

M3DVQ wrote:I would be very interested to know how those paint matching scanners work too. It's possible they use the same principle as the "toy" ones but with more accurate and calibrated optoelectonics.


Yes that is my assumption too. It may well be that there is no low priced way to do it.

Having said that I have suddenly remembered that a couple of years ago I noticed that one of the paint manufacturers was selling such a device - I can't remember the details or price but I am pretty sure it was for "domestic" use - time for more trolling.
gary
 

Postby gary » Thu Apr 18, 2013 10:07 am

aha! THIS is what I would like:

http://www.youtube.com/watch?feature=pl ... 6udrI1cJbE

alas at $767 it is beyond my moderate pocket - but there may be cheaper versions with less "features" I suppose.

I think something like this would be useful even for people with good colour sensitivity.
gary
 

Postby gary » Thu Apr 18, 2013 10:12 am

Here is another one and if you read the description of how it works it sounds so easy - but then they don't give a price and you have to request a quote - that doesn't bode well....

http://www.ferret.com.au/c/IDM-Instrume ... nt-n681700


I am getting the feeling though, that if one knew what to look for, that there are specialised optoelectronic components out there that may not be too expensive.
gary
 

Postby gary » Thu Apr 18, 2013 11:33 am

Here is how one type of analyser works:
http://www.youtube.com/watch?v=pxC6F7bK8CU
gary
 

Postby M3DVQ » Thu Apr 18, 2013 11:48 am

gary wrote:Here is how one type of analyser works:
http://www.youtube.com/watch?v=pxC6F7bK8CU


That would certainly be the way to do it if you want it to "see" all colours correctly. Either diffraction like that, or with a prism. The important thing is having a truly white source, and that's often the hard part!

My instinct is that any reasonably good detector should work just so long as it detects all visible frequencies to some extent. It doesn't matter how non-linear the response is since you can calibrate it with known samples to determine the the response curve and then correct the readings accordingly, just like calibrating white balance really.
M3DVQ
Just nod and pretend you understand me
 
Posts: 338
Joined: Sat Feb 10, 2007 10:52 am
Location: Lincolnshire

Postby Klaas Robers » Fri Apr 19, 2013 6:15 pm

I think that is not too difficult. If you remember how the spectrometer works, with the moving grating or prism, and the detector, if you use a halogen incandescent lamp, then you might calibrate the system by measuring the light of the lamp without any sample in the aparatus. Then you know the spectrum of the lamp itself.

Incandescent lamps have a continuous spectrum, because it is a thermal radiator, just like the sun. A hot body. Only the emitted amount of power is not equal in all wavelengths, this is because of the limited temperature. The higher the temperature, the more short wavelength (blue) power in the spectrum. But it remains continuous, no spectrum lines.

Then when you insert a sample you measure the attenuation of that sample as a function of the wavelength, so the damping of the known / calibrated light spectrum.

If you really need a "paint" that matches with the original paint under all circumstances, then this is the only way. All systems with three filters, red green and blue, and three detectors, give though R G B values, but if you change the light source, you don't know what happens with the colour. And even worse, if you have two colours that match in R, G and B of the "meter", then still you, ore somebody else, might see a difference. No two people see the same colours, as also our eyes work with R, G and B sensitive sensors, but the curves differ from person to person.

Sometimes the R and G sensors have the same, or almost the same sensitivity curve, and then that person sees no difference between red and green.
User avatar
Klaas Robers
"Gomez!", "Oh Morticia."
 
Posts: 1656
Joined: Wed Jan 24, 2007 8:42 pm
Location: Valkenswaard, the Netherlands

Postby DrZarkov » Fri Apr 19, 2013 8:58 pm

In printshops they use colour calibrating systems to get exactly the same colour on the monitor as on paper. This company is selling affordable devices: http://spyder.datacolor.com/portfolio-v ... r4express/

Maybe you are lucky to find a cheaper version: Many printshops throw them away together with old CRT monitors and old Macintosh hardware.

I think this kind of hardware could be what you are looking for?
User avatar
DrZarkov
I think I've had a cranial implosion.
 
Posts: 1041
Joined: Mon Feb 19, 2007 11:28 pm
Location: Kamp-Lintfort, Germany

Postby gary » Fri Apr 19, 2013 11:15 pm

Thanks Klass, that pretty much reinforces my thoughts. The only thing I would add is that I am not interested in convincing other people they are right - only that the colour is right - I would add to that that my own opinions are not only worthless but irrelevant.

I should also add though that very few people are so colour blind that they can't distinguish red from green and that they only have problems with certain shades of each which when looked at on a spectrum are very close together in frequency.

DrZarkov yes that is the sort of thing I am looking for but not for monitors - it is no use to me that my monitor displays correctly, but if the device works out what the RGB is of any object that is another matter.
gary
 

Postby tubefriend » Mon Apr 29, 2013 6:44 pm

If you can bring the surfaces to be measured in very close proximity to an sensor head i would have an idea to measure colors with few cheap parts.

I would build an lamp with an wide range of LED colors, from deep red over bright red, orange, yellow, yellowish green, true green, blueish green, bright blue and deep blue. When you switch sequentially through these LEDs and measure the current of an photo diode you would get an set of values representing the reflectivity of the surface as an function of the wavelength.

Of course this is only relative to your equipment and not an calibrated spectral response.

By using more, for example nine, channels instead of only RGB you catch differences between paints that look similar in one kind of light with one pair of eyes but different in others. The usage would be that you would scan the color that you need, than compare the values with an library of the colors you have.

If you find one where the relative values match well you only need to calculate or try out the needed dilution to match the color intensity.

If you dont find any color that matches well you can search for combinations of colors that can be mixed for the right result, for example an yellow and an deep red if you need some kind of dull orange.

If someone is good in programming he could automatise this in a way that an database knows all colors and tries all possible combinations, then prints possibly combinations that you can try.
tubefriend
"Fester, enough of the light-bulbs!"
 
Posts: 8
Joined: Sat Mar 16, 2013 4:19 am
Location: Germany

Postby M3DVQ » Mon Apr 29, 2013 8:45 pm

Mixing two LEDs is pointless because you will always get the same result as measuring first with one, then the other. Our eyes are fooled into thinking there are new colours but there aren't.

Adding more different LEDs reduces the errors but there just aren't enough different colours of LED.
M3DVQ
Just nod and pretend you understand me
 
Posts: 338
Joined: Sat Feb 10, 2007 10:52 am
Location: Lincolnshire

Postby tubefriend » Fri May 24, 2013 11:52 pm

I meant not mixing two LEDs, but mixing two or more paints. English is not my native language so some points might have got lost in translation and i try to explain it again.

My idea would have been to use as much different LED colors as possible, switched one after eatch other to measure the "Darkness" of the paint at the actual wavelength compared to an plain white. Something like an flying spot scanner with only one pixel but as much colors as possible.

This would not allow true spectral analysis, but at least allow 9 wavelength points with commonly available LEDs and should be much more accurate than an reduction to only R G B.

If you have good luck you will find a Paint in your collection that has the same relative values for all wavelengths, then you can say for sure that it would look the same as the original under any lighting situation.

If you dont have this luck you could try to mix different paints until to get the most similar result. You could for example mix an deep red paint with an orange paint to get something that matches much better than to use only an medium red paint.
tubefriend
"Fester, enough of the light-bulbs!"
 
Posts: 8
Joined: Sat Mar 16, 2013 4:19 am
Location: Germany

Postby gary » Sat May 25, 2013 10:47 am

Whilst I can see how the extra colours could aid in the colour sensitivity of the device wouldn't calibration become a bit of a nightmare?
gary
 

Previous

Return to Off Topic

Who is online

Users browsing this forum: No registered users and 23 guests