Page 1 of 2

True Colour Recognition

PostPosted: Tue Apr 16, 2013 11:06 am
by gary
I wonder if anyone has experience with, knowledge of, or thoughts on, true colour recognition systems? By colour recognition I mean the recognition of colour, not objects by colour. The only example I can think of is possibly the machines at paint shops, hardware stores, etc that you take a flake of paint to. These analyse the paint flake and give the shop man the tints required to make up a matching can of paint.

I am wondering if there are any affordable systems available or whether it is feasible to construct a system.

In either case I would expect that the biggest problem would be calibration - any comments on that would be welcome.

Issues in constructing one would be things like, providing a pure white light source, and then, I suppose have three primary colour sensors with very narrow bandwidths. I don't know maybe there are other techniques? Obviously general "off the shelf" components are unlikely to "cut it", and, of course, calibration seems as if it would be a very difficult problem to solve.

The output of this system would most probably be an RGB value.

I have, of course, seen "toy" implementations of such system but they don't produce a reliable output (I don't think).

Obviously ccd cameras are notoriously bad at colour reproduction.

A Google search at this stage has not been very helpful - "industrial strength" systems are available but I haven't come across any "consumer level" systems.

PostPosted: Tue Apr 16, 2013 3:11 pm
by Steve Anderson
This could be viewed as 'industrial' as I haven't seen it applied to consumer level gear.

In broadcasting colour matching of the banks of monitors is important, the director/producer needs accurate colour rendition on all monitors, "Is that monitor's colour correct, or is it this one?" This also applies to Gamma and black/white intensities.

As far as I can recall Barco were the first to add auto-setup to their Grade 1 colour CRT monitors, some time in the early 80s. This consisted of a rubber suck-on device with three colour sensors inside screened from ambient light. The flying lead plugged into the front of the monitor and you simply pressed the "Auto-Calibrate" button and the micro inside set it all up. You only needed one of these devices to set up and maintain a complete TV station's monitors...assuming they used Grade 1 Barco monitors exclusively...rare...these cost in the region of US$30,000 each at 80s prices.

Sony introduced a virtually identical system later. The only other company I have seen develop such a device is Eizo for their top-of-the-line graphics workstations, both for CRTs and flat-panels.

Steve A.

I know not of much help. but a bit of background...

PostPosted: Tue Apr 16, 2013 3:24 pm
by gary
If I hadn't retired I could have borrowed one of those ;-)

Hmmm, the "auto-calibration" reference triggered thought in my mind that perhaps calibration is not so difficult after all. One would only need a set of those colour matching cards you can get at paint outlets. I know of at least one brand of paint that has the RGB values on the back of the card.

PostPosted: Tue Apr 16, 2013 6:12 pm
by Viewmaster
I think that you need to borrow from my collection (for a small fee :) )
a 'flicker photometer.'

When we went to the 'flicks' many, many years ago this instrument was used to compare colour and brightness.

A clockwork motor rotates a white reflecting tapered plaster disk between two lens systems. The start of the tsper is shown in one of the pics.

One lens is 'looking' at the standard colour and the other the one to be matched.
Rotation speed is adjustable (see artical).
The eye is very sensitive to flicker, hence this is how this instrument works.
Here's an intersting piece on the flicker photometer........

PostPosted: Tue Apr 16, 2013 6:29 pm
by gary
That looks a nice piece of kit Albert, alas I can't see a way of getting an RGB out of it unless I had an infinite number of calibration samples... ;-)

PostPosted: Tue Apr 16, 2013 6:51 pm
by Viewmaster
Could this be any very slight basis for what you are looking for, Gary? ... 037wt_1145

There are documents listed too about how it works.

If you google 'color recognition' (spelt the USA way) you may find something too.

PostPosted: Tue Apr 16, 2013 7:04 pm
by gary
Indeed Albert it is what I am looking for except I am looking for one that actually works ;-) This, I think, is one of the many "toys" I have come across that do a fair job of detecting reds blues greens yellow - but not the tints one needs to make a colour match for paint etc.

Yes Albert, "colour recognition" is my key phrase for my google searches on this topic - I leave the quotes out of course so Google can match colour to color - it's pretty good at doing that sort of thing.

PostPosted: Tue Apr 16, 2013 7:26 pm
by Viewmaster
gary wrote:Indeed Albert it is what I am looking for except I am looking for one that actually works ;-) This, I think, is one of the many "toys" I have come across that do a fair job of detecting reds blues greens yellow - but not the tints one needs to make a colour match for paint etc.


Could I ask why you need to get true colour recognition anyway?....are you going into the paint trade? :wink:

PostPosted: Tue Apr 16, 2013 9:17 pm
by gary
Matching up paints etc is definitely a good use for such a device, and as I build furniture etc for a hobby (with a few small commissions and love jobs) it would certainly come in useful if I need to match various finishes. But I use that as an example that, well, nearly everyone can appreciate.

In truth, if you are red-green colour blind like I am there are a multitude of uses - perhaps even reading resistor colour codes when the multimeter is in use elsewhere or otherwise out of action?

Some lucky people have a colour recognition system at home - it's called a woman - a device that suffers very rarely from colour blindness and generally has colour sensitivity well beyond that of mortal man, alas I do not. (I once used one to help me sort out a box full of mixed up resistor packs so I know how useful they can be - sorry Karen, I know it sounds misogynistic but believe me it is isn't).

But please don't belabour the "what for" point - there is nothing more annoying on a forum when someone asks a questions and some one comes along and says "what on earth do you want to do that for?". And yes I know I am guilty of it myself... ;-)

But if you want a solid reason let's just say I wish to further reduce my dependency on other people and organisations.

PostPosted: Tue Apr 16, 2013 11:22 pm
by Steve Anderson
Now one would think that resistor colour codes would be quite simple, alas, not so...I'm only using this as a base example....

On yer general 1/4W resistoid the bands are 1mm or less wide, the background body colour can bleed through too, usually a blue-ish grey on 1% parts or a cream/beige on 5% versions. Although OK on primaries and their compliments I am slightly colour blind to pastels, someone will tell me those shoes are a brown or tan, whereas to me they are grey...or vice-versa.

What doesn't help is when (often) the tolerance band isn't spaced apart from the actual value, e.g. with 1% parts Brown, Red, Brown, Brown, Brown. (1k2, 1%) or Brown, Brown, Brown, Red, Brown. (11k, 1%). A very common value with potential for confusion is 10k.

Even though fairly confident with being able to read resistor values I do still sometimes check them, even though my stocks are bagged in storage by value, there have been instances where the local supplier has made errors and I haven't picked them up.

The multimeter is of course the ultimate check, but it is a pain having to use it when it shouldn't be required.

We can be grateful I guess that those capacitors which looked like Liquorish Allsorts and other colour-coding systems seem to have fallen out of favour...a printed 102 (1000pF, 1nF) in black text is far preferable...even if I do need a magnifying glass to see it!

Steve A.

PostPosted: Wed Apr 17, 2013 10:53 am
by gary
Yes indeed Steve, as you say there are many things that modify how well anyone, never mind a colour blind person (every one is colour blind to some degree - there aren't an EXACT number of cones or colour receptors in the human eye), sees colour - if the reflected light is not pure white, if it's luminosity is low, background, proximity to other colours etc. Why anyone ever thought resistor colour codes were a good idea I'll never know ;-)

It amuses me that, when the overwhelming number of colour deficient people are red-green deficient, that those 2 colours so often pop up in matters of safety. Especially when there is probably nearly no one who would fail to distinguish between blue and, say, yellow (or red for that matter - and when you think of the colour spectrum that would be the obvious choice).

Very few "colour blind" people would have trouble recognising any pure primary colours - it's when someone asks "what kind of red" or "what tint of green" or what "shade of blue" that the trouble starts - and that is exactly what I am looking for in a colour recognition system.

PostPosted: Wed Apr 17, 2013 10:45 pm
by AncientBrit
You may have another potential problem regarding the spectral purity of the light source under which sample and reference are compared.

Fluorescent sources are very 'peaky', incandescent has a more even distribution.

Depending on the pigments used in the paint it is possible that a match could be indicated say in daylight, only for a mismatch to occur when viewed in artificial lighting.

Hence shop customers moving to the doorway when they are comparing swatches of material.



PostPosted: Wed Apr 17, 2013 10:50 pm
by gary
Indeed Graham. that's what I meant by " if the reflected light is not pure white" - thanks for the clarification.

PostPosted: Wed Apr 17, 2013 10:53 pm
by AncientBrit
That's what I call a lightning response, 5mins!!


PostPosted: Wed Apr 17, 2013 10:57 pm
by gary
There are "overlap times" between Northern and Southern hemispheres... but it won't be long before this little retiree is in bed ;-)