4 ms·
>the only color system produced entirely from measurements of human color perception What is CIE LAB then? While Wikipedia states that it was influenced by the
by leni536 11y ago
>the only color system produced entirely from measurements of human color perception
What is CIE LAB then? While Wikipedia states that it was influenced by the Munsell color system, I don't think it was directly derived from it. Also I would argue that it is in common use.
On the other hand these measurements measured the local distance between colors (based on distinguishability) but nothing guarantees that the space of perceivable colors is Euclidean in this metric. So you can fabricate an Euclidean space of colors where the Euclidean distance is close to the perception distance, but it may never be perfect.
Side note: there is no perceivable color space to "rule them all", even if you can fabricate a perfect one. For upscaling images it could make sense to use a perception based colorspace for interpolating pixel values, however for downscaling you should use a linear colorspace (note: sRGB is NOT linear, almost all desktop software downscales wrong [2]). It causes kind of a hassle that distinguishing colors of large areas is happening with a strange metric, but if you are looking at objects with fine structure (like your LCD monitor) then the color mixing happens in a linear colorspace.
[1] https://en.wikipedia.org/wiki/Lab_color_space https://en.wikipedia.org/wiki/Lab_color_space
[2] http://www.4p8.com/eric.brasseur/gamma-1.0-or-2.2.png http://www.4p8.com/eric.brasseur/gamma-1.0-or-2.2.png
- icegreentea 11y agoFunny, I just watched a Scipy conference talk/video about this - it's about finding a new default color map for matplotlib [1] CIE LAB is basically a ~30 year old model based on human perception - basically it takes into account white point adjustment and the non-linear human brightness response, and thats about it. There are newer (and faaar more complicated) models like CIECAM02. [1] https://www.youtube.com/watch?v=xAoljeRJ3lU https://www.youtube.com/watch?v=xAoljeRJ3lU
- leni536 11y agoAs I see CIE LAB is not a something "inferior model" to these more complicated models. It just has a more simple viewing conditions model and a more limited regime of validity. It seems somewhat complicated to properly apply CIECAM02. In the case of matplotlib, one should change the line colors depending on the background color being used. Not talking about adapted whitepoint and surrounding too where you at best can make an educated guess. They seem to solve the Euclidean distance problem by embedding in a field with more dimensions. However I'm really happy for these efforts of matplotlib. It is by far my favorite data visualization library. There could be cases where this kind of accuracy is necessary: doctors often diagnose using visual inspection of different camera images (X-Ray, CT, MR). The optimal visualization of these camera images using well crafted colormaps in the right colorspace could save lives.
- mrgriscom 11y agoMy understanding of CIELAB is that it tries to approximate perceptual uniformity while also having a rather simple transformation formula. As such the perceptual uniformity is compromised (intentionally). It is a good approximation but the lack of uniformity can be noticeable -- particularly in the hue dimension. See my 3rd link.