Can I Use Additional Wavelengths for a More Advanced Hue Calculation?

[Physics_Dave]

TL;DR Summary

When calculating Hue values, are we limited to just RGB data?

Hello,

I understand we can calculate a Hue value from RGB data. I was wondering if we are limited to red, green and blue for that calculation. If I also have data from additional wavelengths (Infrared, Far Red, Amber, UV, and White), can I use the additional intensity results to create a more "advanced/in-depth" hue Calculation?

For context, I am using a Keyence MulltiSpectrum vision system where you can use up to 8 different color lighting wavelengths on a monochrome camera, to create a color image. This setup allows for not just RGB data, but all the additional lighting wavelengths as well.

Thank you!

 

[pinball1970]

TL;DR Summary: When calculating Hue values, are we limited to just RGB data?

Hello,

I understand we can calculate a Hue value from RGB data. I was wondering if we are limited to red, green and blue for that calculation. If I also have data from additional wavelengths (Infrared, Far Red, Amber, UV, and White), can I use the additional intensity results to create a more "advanced/in-depth" hue Calculation?

For context, I am using a Keyence MulltiSpectrum vision system where you can use up to 8 different color lighting wavelengths on a monochrome camera, to create a color image. This setup allows for not just RGB data, but all the additional lighting wavelengths as well.

Thank you!

Hi Dave,
UV and IR are not seen by the human eye as they are outside the 400-700nm range so would not actually add anything IMO.
White light is a combination of many "visible" wavelengths and those outside, so will add.
Is this for graphics?

 

[Physics_Dave]

Hi Dave,
UV and IR are not seen by the human eye as they are outside the 400-700nm range so would not actually add anything IMO.
White light is a combination of many "visible" wavelengths and those outside, so will add.
Is this for graphics?

This is a check to inspect for color of anodized metal parts, before they leave a manufacturing facility. The spec specifically calls out hue values to determine pass or fail, but was hoping to utilize as much data as possible to go into that calculation. Figured if I had additional wavelengths data outside RGB, it could make the calculation even more robust.

That may not be the case... hmmm

 

[berkeman]

This is a check to inspect for color of anodized metal parts, before they leave a manufacturing facility. The spec specifically calls out hue values to determine pass or fail,

A good spec would generally define the test method, or list the instrument(s) that are acceptable for performing the checks. How specific is this specification?

 

[Dale]

This is a check to inspect for color of anodized metal parts, before they leave a manufacturing facility. The spec specifically calls out hue values to determine pass or fail, but was hoping to utilize as much data as possible to go into that calculation. Figured if I had additional wavelengths data outside RGB, it could make the calculation even more robust.

That may not be the case... hmmm

If you calculate a “hue” based on wavelengths that the human eye cannot see then two parts may appear the same hue to the human eye even though the calculated numbers are different. Conversely, two parts that appear different to the human eye may have the same calculated number.

Is that desired?

 

[sophiecentaur]

TL;DR Summary: When calculating Hue values, are we limited to just RGB data?

I understand we can calculate a Hue value from RGB data.

CIE colour space can be represented on the CIE colour chart. The perceived colour of a source can be plotted on the graph below and its xy co ordinates are arrived at by the relative outputs (normalised to eliminate luminance) of the standard human colour analysis (filters on the retinal sensors). A 'blue' hue will be somewhere around (0.15,0.1), for instance.

In normal colour TV, they use three 'primaries' and any hue that lies within the triangle on the diagram can be produced by adding suitable weights of the primaries. Source colours outside the triangle will appear somewhere along the sides of the triangle. Other primaries can be added, say in the region around the greens / cyans and the R signal would be kept at zero to allow the extra phosphor (say at (0.1,0.7)) to 'pull' the resultant outside the triangle. If you want some more details about the calculations, this link could be a start.
In practice, there is little to be gained from adding extra primaries for a tV display because extra phosphors would take up room on the face of the display, limiting the available brilliance or the resolution. The spectra of phosphors are chosen not to be monochromatic but for efficiency.

 

[pinball1970]

This is a check to inspect for color of anodized metal parts, before they leave a manufacturing facility. The spec specifically calls out hue values to determine pass or fail, but was hoping to utilize as much data as possible to go into that calculation. Figured if I had additional wavelengths data outside RGB, it could make the calculation even more robust.

That may not be the case... hmmm

Ok those wavelengths we cannot see but are there should add to the measurement.
I measure surfaces between 360-750nm but I obviously cannot see that range, my spectro can though.
The standard is stored, a trial measured against it, then values compared at each wavelength, a pass fail is then generated using the reflectance data. Standard red and a trial in blue.

How do you measure the surface?

 

[sophiecentaur]

Ok those wavelengths we cannot see but are there should add to the measurement.

What you see is essentially a colour and not 'a wavelength', as your spectrometer will probably show you.

Ok those wavelengths we cannot see but are there should add to the measurement.

The non-visible wavelengths that come off the surface will not affect the colour and, if you are colour matching, it's just a red herring.

One good reason for using spectrometry can be variations in illuminant. A metemeric match (subjective) can be obtained with a whole range of pigments but they all go wrong if you switch the lighting. Matching a shirt in one shop with a jacket in another can often fail.

If you are using lab conditions then digital photography (same illuminant all the time) can be useful and your pass / fail system could be quick and cheap, with only three numbers involved (RGB)

 

[Andy Resnick]

This is a check to inspect for color of anodized metal parts, before they leave a manufacturing facility. The spec specifically calls out hue values to determine pass or fail, but was hoping to utilize as much data as possible to go into that calculation. Figured if I had additional wavelengths data outside RGB, it could make the calculation even more robust.

That may not be the case... hmmm

I agree with @berkeman - what does the spec say? Is there a reference to one or more ASTM standards?

 

[pinball1970]

What you see is essentially a colour and not 'a wavelength', as your spectrometer will probably show you.The non-visible wavelengths that come off the surface will not affect the colour and, if you are colour matching, it's just a red herring.

One good reason for using spectrometry can be variations in illuminant. A metemeric match (subjective) can be obtained with a whole range of pigments but they all go wrong if you switch the lighting. Matching a shirt in one shop with a jacket in another can often fail.

If you are using lab conditions then digital photography (same illuminant all the time) can be useful and your pass / fail system could be quick and cheap, with only three numbers involved (RGB)

So the wave lengths do not contribute to the measurement at all?
I can discount the first and last four values?
Why that range on the spec?

 

[sophiecentaur]

So the wave lengths do not contribute to the measurement at all?

If the application is getting a good, consistent match with human vision then invisible wavelength measurements can only detract from the results.
EDIT: what does that remark mean? It's the total effect of all the wavelengths that affects the subjective colour. If there are more wavelengths present than one then the 'colour' is not a wavelength. We very seldom actually see monochromatic light. Virtually all visual experiences involve a wide range of wavelengths. most of the 'vivid' colours we see are a long way away from that spectral curve on the CIE diagram - take a look at that triangle of primaries it's miles away from spectral greens yet we like the TV pictures.

Why that range on the spec?

You'd need to speak to the person who wrote the spec. Is there some other use for knowing the invisible spectrum? Could it be to do with monitoring the anodising processing? I only know about human colorimetry.

 

[pinball1970]

If the application is getting a good, consistent match with human vision then invisible wavelength measurements can only detract from the results.
EDIT: what does that remark mean? It's the total effect of all the wavelengths that affects the subjective colour. If there are more wavelengths present than one then the 'colour' is not a wavelength. We very seldom actually see monochromatic light. Virtually all visual experiences involve a wide range of wavelengths. most of the 'vivid' colours we see are a long way away from that spectral curve on the CIE diagram - take a look at that triangle of primaries it's miles away from spectral greens yet we like the TV pictures.

You'd need to speak to the person who wrote the spec. Is there some other use for knowing the invisible spectrum? Could it be to do with monitoring the anodising processing? I only know about human colorimetry.

I meant, do those values either side of 400-700nm contribute nothing to the overall measurements that go into the Delta E pass fail?

My previous instrument measured between 400-700nm so I was surprised when we acquired this instrument and saw the range. I assumed it was to get the data on the edges more accurately rather than a complete cut off. I will ask the tech who does our annual service.
Knowing what is there rather than just what we can see is useful in terms of chemical composition possibly? Like the amount of oxidized metal on a surface.

Need to know from @Physics_Dave :

Instrument he is using to measure
Equation
Tolerances for pass for pass fail

 

[sophiecentaur]

I meant, do those values either side of 400-700nm contribute nothing to the overall measurements that go into the Delta E pass fail?

I had no idea about Delta E until I did some searching. I suggest you read this link and anything else you can find. It's based on a test, using the three Tristimulus filters which model typical human colour vision. There's a formula in that link which tells how they arrive at a score.
It's far easier to get a realistic Delta E score for a monitor than from a printed colour because the appearance of a colour, reflected from a surface will, as I pointed out earlier, will be strongly affected by the illuminant. Otoh, the light from a monitor screen comes direct.

There is an HSV standard which works better than the simple three colour measurements. It uses six colours instead of three, but I reckon it's only worth while for very critical colour matching (like getting right the Coca Cola Red colour which we all know very well).

 

[pinball1970]

I had no idea about Delta E until I did some searching. I suggest you read this link and anything else you can find. It's based on a test, using the three Tristimulus filters which model typical human colour vision. There's a formula in that link which tells how they arrive at a score.
It's far easier to get a realistic Delta E score for a monitor than from a printed colour because the appearance of a colour, reflected from a surface will, as I pointed out earlier, will be strongly affected by the illuminant. Otoh, the light from a monitor screen comes direct.

There is an HSV standard which works better than the simple three colour measurements. It uses six colours instead of three, but I reckon it's only worth while for very critical colour matching (like getting right the Coca Cola Red colour which we all know very well).

DE is good for surfaces. Spectral data to tristimulus to Lab to DE.
The software does all the fancy stuff, I just get DL DC DH and overall DE pass fail.
What you see usually aligns with what you see in the light box unless there is some sort of FWA or fluorescent dye.
The software is pretty nifty as you can stick all sorts of metrics in there, Metamerism index, absolute Hue angle. I'll send a screen grab when back in the office.

Anyway, @Physics_Dave ? How do you measure?

 

[sophiecentaur]

What you see usually aligns with what you see in the light box

How is the illuminant specified? Tungsten at a given temperature?

 

[pinball1970]

C.I.E. (https://en.wikipedia.org/wiki/International_Commission_on_Illumination) specify the illuminants, the SPD and that goes into the software for pass fail measurements.

Then we can select whatever primary light we want, depends on where the object will be viewed eventually.

If it is a car, then CWF used to be primary as that is what show rooms had.

Due to energy efficiency LEDs are being used now in a lot of commercial settings.

LEDs will save a lot of energy but the SPD is weird, I am still getting used to the effects.Tungsten is pretty much phased out, I still have the bulbs for the boxes but I do not use them.

Energetically very inefficient and poor colour rendering.

I measured my black bag today and selected D65 as primary, CWF and an LED then measured a different part of the bag for a trial.

The software gives me the colour data in terms of LCH values and trial has the deviations from those values.

As it is the same sample I expected a pass and got one. Not perfect but less than 0.5DE overall which is
very close.

We could do with @Physics_Dave Dave letting us know his kit/ software and pass/fail tolerances.

 

[sophiecentaur]

Tungsten is pretty much phased out,

But has Colour Temperature been phased out? Light sources which are not from a hot body are still commonly described in terms to temperature. Sloppy, maybe but if you want to improve on that then you should define the spectrum of the source.

Energetically very inefficient and poor colour rendering.

Absolutely, very inefficient but very reproducible so I think that statement could be putting the cart before the horse. Colour rendering has to include a specification of the illuminant. LEDs vary between themselves and are only reliable sources if they are specified consistently. My only experience of LED lighting is the domestic stuff which is very loosely specified and sold on its light level more than anything else.

C.I.E. (https://en.wikipedia.org/wiki/International_Commission_on_Illumination) specify the illuminants, the SPD and that goes into the software for pass fail measurements.

The only mention of specification of illuminant in that link is D6500 which is a black body at 6500K (a reference Sun). How is the illuminant in your machine specified (if not a hot object then what)? What do you do when your illuminant source dies? It bothers me that you go for a match under the intended viewing conditions but is that useful for clothing which will be viewed in the outdoors (full sun, cloudy and sunset), tungsten (still) and unspecified indoor lighting? There has to be at least a caveat about the likely match and what's the most and least suitable lighting - I'm talking about fashion, in particular where people need plastic, knitted and woven fabrics in an outfit and items which come from many different sources.

I guess you'll tell me that your machine works fine in the context you use it - and that, of course, is totally fair.

 

[pinball1970]

But has Colour Temperature been phased out? Light sources which are not from a hot body are still commonly described in terms to temperature. Sloppy, maybe but if you want to improve on that then you should define the spectrum of the source.

Absolutely, very inefficient but very reproducible so I think that statement could be putting the cart before the horse. Colour rendering has to include a specification of the illuminant. LEDs vary between themselves and are only reliable sources if they are specified consistently. My only experience of LED lighting is the domestic stuff which is very loosely specified and sold on its light level more than anything else.

The only mention of specification of illuminant in that link is D6500 which is a black body at 6500K (a reference Sun). How is the illuminant in your machine specified (if not a hot object then what)? What do you do when your illuminant source dies? It bothers me that you go for a match under the intended viewing conditions but is that useful for clothing which will be viewed in the outdoors (full sun, cloudy and sunset), tungsten (still) and unspecified indoor lighting? There has to be at least a caveat about the likely match and what's the most and least suitable lighting - I'm talking about fashion, in particular where people need plastic, knitted and woven fabrics in an outfit and items which come from many different sources.

I guess you'll tell me that your machine works fine in the context you use it - and that, of course, is totally fair.

I really hope the OP will come back and extend his 4 hours and 16 minutes participation on PF ...the mods will allow this slight digression till he returns?

In terms of light specification for CACs organizations like CIE define the light in terms of the SPD and this is filtered down to the buyers, retailers with help of organisations like the SDC and manufacturers of the spectros and light boxes.
As I mentioned the lights relate to point of sale so the science relates to the industry.

The light box people include CCT and lux in the spec documentation and I know at least one organisation that gives the SPD curve but not the data.
So I can buy a box and check with a radiometer that the output and colour are close to the spec.
The spectroscopy guys have all the SPD data already in the software. So I can look at a measurement without going anywhere near a CAC or sample and know what it looks like in all relevant lights and spot potential issues.
Your question about D65, yes that is intended to mimic daylight (Northern hemisphere, over cast , midday)
So two challenges we want a match in that primary light D65 as it is summer wear BUT it is being sold in a store that has LED with a 4000 CCT.
So our dyers need a combination that will achieve the colour in the primary light, store light AND meet fastness requirements.
Short answer no.
The chromophore will do something different under a different SPD, no avoiding it. You try for a happy in-between.
It will change but not too much

 

[pinball1970]

Too many acronyms.

C.I.E. 1900 organisation that makes specifications on lights. Based in Belgium.
SDC. Society of dyers and colourists, based in Bradford York's, UK.
CAC. Colour assessment Cabinet. Different ones depending on the end use. Will have standard light sources like D65, TL84, Tungsten and UV. Newer ones have LED.
CCT. Correlated Colour Temperature.
SPD. Spectral power distribution.

 

[sophiecentaur]

So our dyers need a combination that will achieve the colour in the primary light, store light AND meet fastness requirements.

That's a good engineering approach. People will always be using CFLs (sorry) and they deserve to look unwell around their homes.