 Hello, and so a little bit delayed, but I kind of start my presentation after having achieved HDMI input to the projector and My name is Evin Koalos. I'm involved in the GIMP project and And my initial education was in fine art I went to art school in a place in Spain called Cancerat where we last year had Wilbur week as we called it week-long meeting of the GIMP developers on a Masia in Catalonia and And when it comes to color, which is a private thing this presentation is about even before I went to art school I was rather deeply immersed in it because my father was a machine repair technician for Norway's largest paint manufacturer who Offers this wide spread of colors available for all interior paint colors that you would use inside your house But these were not colors that you they wouldn't have One liter and three liter and ten liter buckets of all those quite a few thousand colors In each color shop in paint shops, they would have a set of I believe it was six base colors and then they had pure pigments in a Water solution that got mixed in and they had recipe books of how to generate all these colors It's similar to a Pantone kind of swatch system whereas these fans are Another system called the natural color system, which is common in Scandinavia In Germany, there's another system which has used on some metallic paints from the same manufacturer This is called the roll, which is common for such color matching So I grew up since I was nine ten years old in a household where there are rather hardcore color signs teaching books and similar and The actual thing I want to talk about today is how you can match colors for printing and This is book for painters printed in 1725 about How as a painter you can Reduce coloring to mechanical practice what we would call an algorithm today So it is a book that Has color illustrations This is an example of a painter's color palette in a book printed in 1725 And this is an example of the pictures that this J.C. Leblanc Has so this is a full color the result he wanted to illustrate and The illustrations in this book they actually are explaining for a painter how to do things but the interesting thing is that he actually had full color illustrations of both of both the palette and this thing and and What he actually did was that he invented four color printing basically and He didn't use Simon Gent and yellow. He used red blue and yellow and Also added black for Additional color and and It is one of the kind of earlier of kind of painters color science books that actually kind of gets things much more, right? he praises Isaac Newton if you see the last line about paragraph and And he understood the differences between additive and subtractive color mixing Whereas even when I went to art school back in the late 90s the color science We got which is items color science is not really how color works, but it's Thinking about how it might be working so Isaac Newton Made a discovery that when you pass light through a prism it splits into colors and What happens when this light goes into the prison is that we take energy distribution on many different frequencies and we spread it apart we get the rainbow and the visible part of the electromagnetic spectrum is only a small part and On the sides of the visual Parts we have ultraviolet light and infrared light, but also microwaves and radio waves and And X-rays are part of the exact same physical phenomenon It's just that our visual system is only attuned to a small part of it other animals are attuned to other parts and To show yet another kind of illustration on top of the spectrum I'm going to be showing quite a few this type illustrations This is the energy distribution of daylight In the northern hemisphere is also known as the d65 illuminant So this is how the combination of like the Sun and the Sky which illuminates things when we are outdoors. This is the total amount of energy there is and An interesting question is what is color at all and Quite a few people they come to the conclusion that color is a subjective experience and it only exists in the minds of humans and and After having gone to art school and up in the University in Norway that just around that time was starting up in the region color laboratory and There Focus was on printing and Something that is known as colorimetry and colorimetry is the study of color as it is perceived by humans and One central part of that is this and this shows which part of the visible range of light We are sensitive to in that perceiving it as luminance, but I co-bride something is and It's in the middle of the spectrum where Small amount of energy would give the same as a high amount of energy On lower or higher wavelengths and But the way our vision system works is also quite strange and interesting because These spectrums we get we don't actually really sense them we we get some Three-dimensional value kind of like almost like RGB ends up going into our visual cortex these illustrations show the spectral distribution of primaries of different types of displays and monitors and And if you have a laser-based projector, you have this very narrow spikes on very specific wavelengths whereas CRT screens with phosphors you have this much wider spread functions and And in the end you can show images on these screens and you calibrate the screens and you get the perception that they are the same and in studying how and how to achieve that consistency and We've been or humans have been studying on all our own visual visual system since the early 1900s it kind of in detail and in 1931 a Standard called CIE XY set was developed and This was done by studying how using a Red light green light in the blue light Ideally with very narrow spectrums similar to the laser projector If it was possible to mix the colors of the rainbow so one would start out at 400 nanometers and Generate a spectral light monochromatic light of that blue color who knows the spectrum distribution and then then we'd have people trying to use faders to adjust a Red light a green light and a blue light and match the color you have and One went through the entire spectrum to create such matches It turns out that there are some colors that one could not match but one had those three red green and blue and But it was a workaround a hack one could use when you use the split screen And you had the spectrum color on one side and then you were mixing red green and blue lights But then you couldn't achieve the color you allowed to add colors also on the side with the spectral light source and That was considered to be a negative amount of light to achieve the mixture So this graph shows the CIE RGB 1931 color space. So that is the Each of these curves show how high each of those faders for the red green and blue lights needed to be to match each wavelength and Both the red and the green and the blue go slightly below for some wavelength and People didn't like this. This is 1931. We don't have computers Computations are done with slide rules and negative numbers are difficult So what create the distance that so this is CIE RGB and this is CIE XYZ and what one did to create this space is that one decided that it would be nice if One of the components of the color space was exactly the same as this So this is kind of like how sensitive whole luminous we consider something to be so they took the three-dimensional RGB space and Rotated and scaled it using a fine transform until they achieved these response curves instead and if you integrate an Emission spectra or a reflectance of a surface Over these curves the result you get out is XYZ This is yet another similar RGB space and But this is the short medium and long wavelength response functions and these response to the Kind of RGB colors which are on the retina of the eye Neither these functions nor these actually have it do the retina they have to do those color matching experiments So these are kind of the spectral sensitivities of the retina This is not sufficient for doing the full model of the human visual system because the lens Also changes the color you perceive and the color of the lens even changes as we age So the color perception is actually gradually changing as a human gets older Quite a few years ago I present this more lightning talk at LGM where I was playing around with spectral mixing of colors and That was the thing I wanted to talk about here today as well and For me, this is kind of one thing I've been playing with in kind of returning to play around with drawing and art things This is a drawing where I've done a cheap separation to Cyanish blue green and black and generated This separation plates and then I've drawn those separation plates by hand in with cross hatching And this is another example of an image where I've used red and green ink and the fun thing with red and green in the subtractive mixing is that red and green and basically becomes black so You get a quite wide range of colors from just mixing two separate colors Me when you do mixing of just lights which you're pointing and That you do on computers or GB you have what you call additive color And if you were to do additive color with spectra, you would just sum them together. So You have what we actually have here is a blue at the top and a yellow and light source But this is actually how it ends up in subtractive mixing and in subtractive mixing You take a vague wavelengths and color when you mix so the more you mix you go closer to black more more get absorbed What could even imagine that this curtain? And in some sense as that the color only exists in our mind But if you will say that this actually had the color it would be every color with this Burgundy red color here because the light hits it and the light that sticks into it are only the wavelengths which are a reddish and Or rather the only thing was reflected are the burgundy reddish ones. So the curtain itself and Is all the other colors because it absorbs those wavelengths. So I'm gonna show a demo of this separation engine and Illustrates kind of get some custom separations done on the fly. So I have a little scripting language and This spectral color simulation I Initially call them inks I had there but that's not strictly true because I try to actually model how both ink or paints or similar types of coats would behave So I can both specify an RGB triplet or directly a color and Or I could even write out a full spectrum saying starting at this wavelength with gaps of this size and then list all the spectral measurements and Well, I'm gonna replace that with cyan dark magenta, these are spectral measurements that have been done and yellow and it It doesn't get to the full RGB image back it gets slightly Dissaturated and I was quite happy when that ended up being the case because that means I'm doing things quite right but more interesting might be to Not go side magenta, then I failed to actually print properly But to do the example I mentioned to have red and green and To me that is a quite good result for red and green. I would like to ask are there any color deficient people in the room? Color blind color deficient Normally when there's many designers, there's some of them at least amongst open-source designers And not going to mention names But for people who actually Are color deficient The most common one there the red green type of color deficiency. They find this to be Nice color image if you use yellow or orange ish and blue for mixing as to Complementary base colors and the pellets you can create with those which is quite wide range of pellets as well Those they say yeah, that's colorful, but if you do the thing I did first with Red and green I've asked this to someone actually is color deficient and he says that just looks a very desaturated image It's Not good at all, but the blue and yellow blue orange is like yeah, that's a color image So yeah you for Color deficient people it turns out you don't need three or four color printing two might be enough But those observations is also something if we're to actually design pellets for use in something One can kind of use the knowledge that Gradients from the gray axis towards blue and yellow. That is something that 95% or more population can easily perceive So this thing I'm doing here, I'm gonna come back to that the experiments here It's got like my kind of attempt of doing this It's got a debate in pole of learning by doing or in the programmers case learning by coding I don't really understand how something works until I try to implement it and Figure out what about the gaps in my understanding are But things you can do when you Specify colors in this way and you generate more custom things is that one could also imagine to create a t-shirt Which maybe was green so we say we had a green substrate and then we want to have a Print white on it, but a I Have two things there It's just code one equals white and code one black equals and that's RG C O C O C O which is black But I'm gonna say that the first coat and Is white if you draw it on white, but if you draw it on black, it's also going to be white And then I'm gonna add a final layer of black and What this does is that it? generates two inks And and imagine that you have like a green t-shirt 30 than abs white to create the highlights and then create black and I can say that I don't want to separate and proof I just want to oh wait. Let's do separate and proof. So I That is what it will print in white and that is what it will print in black or just a separation as kind of plates it generates those two a starting points for and and This type of Proofing is something that would be useful in GIMP or inkscape or similar applications For simulating printing on such things, but I'm not sure if my code is what one should use there. I'm gonna come back to Developments that have happened since I was playing with this Before the LGM in Leipzig. I Also had one more version of that And separation operation that instead of Doing the inputs as a script only has sliders for doing it. So if we But this is something that doesn't really do the true thing because it Generates the spectral measurements of the color based on the RGB triplet and And at least the way I'm currently doing it I'm almost doing like the color matching experiments. I have a Rather wide band red green and blue components of a scale and I have made sure that I end up with the correct color values out the thing that we use for printing in the world mostly is ICCv4 and that's doing Colorimetric color management and colorimetric color management is only based on these Three-dimensional RGB things and doesn't take the spectral thing into account. So it says that a human perceives it as being the same There's extensions to ICCv4 called ICC Max. ICC Max Goes beyond colorimetry It allows spectral measurements of spot colors But it also allows you for for simulating metallic inks and it allows things like Colors that look different depending on the angle you hold the paper something on its goniometry and And the goal is to do all the things that they know about light and color And push the envelope as far as possible It's a playground for color scientists, but the industry has also started adopting things so the most important thing is CXX CXF X-4 which is an ISO standard for exchanging spot color measurement data and I just want to show my current way of storing colors. I store them the same way that I store the spectrums but it's I have this long Lists of numbers Starting out with the starting nanometers the gap between the spectral bands measured a scaling factor for all of them Not really necessary and then all the spectral measurements CXXF-4 is Exactly the same data, but wrapped up in X-Mouth file format and The first time I stumbled across it. I was quite happy Because I discovered that the measurements that are needed for the full spectral characterization is Printing the color on both white background and black background Which is the kind of the basis of how I was doing my experiments in simulation as well, which means that At some point if I end up using spectral measurements I would be able to feed in the ISO standard way of measuring spot colors into the engine and The full standard goes beyond just having this solid color printed on white and black you also print the a Gradient a ramp of the color to have a transfer response curve and Tell by gamma function or a dot density So you measure measure the spectra of both just a paper and 10% 20% 30% until you have full coverage of ink both on a white background and a black background There is simpler versions of the spectral characterization where you either only have the solid printed on white or Only the ramp on white, but having a full behavior of how the color behaves You need to have almost all these measurements. This is many more measurements than you do with the color of metric isa cv4 But you are able to predict a lot when you do this The thing I compute when I have this information is that I figure out whether the colors behave more like inks or more like paints for each individual wavelength and Then I do one computation for this is how it would be if it was pure inks And this is how it would be if it was paint and then I do a weight between them depending on that computation This is another thing I've been playing with which is Printing roster images with plotters And for those experiments I was also doing very simple color measurements But I didn't kind of pull it far enough to get something else very happy with This is one actually the only color print I've tried I've I'm traveling around too much and I haven't been in the space with my 3d normal 2d drawing machine Since January The code for my engine for playing with this which is quite compact C is on my github and I've decided to call the project for loose flight Spanish for light and On my website there is a gallery of an earlier run of this color simulation project and I'm going to update the results there with how it currently generates results and But I think I should open up for questions maybe It's a you find it both in the Library of Congress in the US and you find it in the French National Library It has been scammed the scammed version in the Library of Congress is bad because Half of the illustrations are missing you have to fold them out And they have just scanned the pages without holding out the illustrations but the one in the National Library of France you can get the PDF of it and it's an interesting book in that Okay, you want the photo of it on the front page It's interesting this is English Okay, you have characters are not used to anymore even in English. It's readable, but it is Bilingual it's printed in both French and in English so it's a book with red yellow blue black four-color printing bilingual from 1725 and This JC Leblanc is basically considered to be the inventor of the four-color process and it's kind of an Oddity that the subject is that he writes a book for painters on kind of like a More proper color theory, but he's doing magic himself and creating full-color illustrations He was doing it in Mesosint. So he was engraving the plates manually in Mesosint, and it's basically equivalent to half toning For the experiments I had to get my own sources and it was measurements I got from a former colleague of mine at the ColourLab in Norway and I also Found some other measurements here and there that I copied in the data for the CI XYZ response functions and similar you find on The international color consortium's own pages you have the measurements But there's also another source of spectral data, which is the Freya Farbe They have those spectral measurements of all of their test prints of LAB. I think they have done ink separation and I think they printed them and then measured them again But yeah, they're spec spectral measurements of those colors, but they for each Sealab or CLCH color. There are many spectral curves, which would give you the same Perception of them. So this is just an example of one at least plausible achievable spectrum that would yield those colors and I think it would be more interesting to Add the ability to Work on arbitrary combinations of colors and There's one thing that I forgot to stick in the slides and that is that this ICC max. I should I jump to a web page What was it called again? Yeah For ICC v4 It was kind of the compromise between people in industry of how much they wanted to collaborate and get to get color management even working And but for ICC max the international color consortium actually has made an open source Project, which also contains an engine for doing Similar things to what I was doing in GIMP Simulation of spot colors and they go much further than what I'm doing. They do include the things about metallic inks and and the Goniometry the angle that you hold the measurement and The goals for them is to be able to do packaging designs say beer cans Where you have a metal surface and you have then inks on top of it and to create 3d model soft proofs So you can see how it changes in the shine as you move it in your hand So you could simulate the whole printing process Without actually doing test runs and test runs and test runs I've looked at it New of this year is that kayuva barman, which some of you know of has added CMake Build files, so it's possible to build it an early nooks now. It used to only build under windows and on the Mac Okay