10/5/2023 0 Comments Color mixing palette![]() But how do we "find" those lighter colors? The problem is that all of those colors are simply too. The problem isn't that we're missing any base colors. But remember, as I stated above, our core palette of paints already gives us a pretty solid representation of every "base" color that we'd expect to find in an image. If we wanted to, we could use the color-mixing algorithms from the previous article to create a whole list of new colors based upon mixing all of the original colors together in different proportions. If we want to perform more accurate color matching against the lighter areas of her skin, the answer is. So although we have a fairly "workable" transformation of the image, matching the pixels in her face to the colors in our paint inventory, there are still areas where this transformation falls short. When you add in additional factors (such as lighting and makeup), their skin can, in certain regions, be incredibly light. But in the image above you can see that many areas of her face are actually quite. They're "dark" - compared to white people. Furthermore, the skin tone of most Black people isn't really all that. But just as "white" people are not truly white (they're mostly a mix of yellow / pinks / tans), most Black people aren't truly black (they're mostly a mix of browns / yellows / tans). it causes problems.įor example, this was the original image we were working from:Ĭlearly, this is a Black woman. But with something as nuanced as a human face? Well. So if you match those images against our palette of heavy body acrylic paints, you may find that the color matching performs quite well. But what this means, in a practical sense, is that all of the off-the-shelf colors are really quite dark.Īnd of course, some images are chock-full of dark colors. The idea is that you can mix them with other paints and they won't immediately lose their core qualities - because the original paints are chock full of pigment. You see, when you buy "professional-grade" paints, those paints come with a very high pigment load. Although we have a broad spectrum of colors, nearly all of those colors are exceedingly dark. for starters, take a look at that palette. So why does our algorithm still fall short when trying to match the digital image to those paints? But there's also reds, and oranges, and yellows, and greens, and blues, and purples, and browns. There's black and white (and numerous shades of grey). If you look carefully at that selection, you'll see that we have all of the "normal" colors one would expect to find in a palette of paints. The reason is that: Paints, especially "professional grade" paints for artists, are fairly dark in nature.įor example: Even though I have 200+ paints in my inventory, here is the color key for all of those paints (you can view this in full resolution here: ): There's a reason why it's inherently difficult to perform color matching when your source image contains a subject like a human face - but your reference palette consists of "stock" paints. Given that I started with an original palette of 200+ paints, why is it that the transformed image has many colors that still don't seem to properly "map" to the source image? As previously mentioned, there's a lot of red/pink in her face where maybe it shouldn't be. But if you step back and blur your eyes a bit, the coloring on her face doesn't look entirely unnatural.īut I'm not entirely satisfied with this image. And overall, you could argue that the image is a bit "noisy". Granted, there are a lot of pinks/reds on her face - a fact that seems a bit odd, considering that she's a woman of color. And the colors on the transformed image are. We don't have any annoying color bands (because we've employed dithering). Finally, I showed how to use dithering to ensure that those "closest" matches were not all bunched together in specific bands of colors.įor reference, this was the original image that we'd pixelated, using a basic RGB algorithm: I then showed how to find the closest match between a given color and a reference palette of colors. In the previous articles, I showed how to pixelate an image so we're not dealing with millions of colors. ![]() Although this can be useful in theory (e.g., to determine what two paints will look like if we mix them together in the real world), this also has practical applications for the live app that I've built at. We can take two colors and "mix" them together to see what the resulting paint will look like. In the last article, I illustrated how we can mix paints virtually.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |