How do we measure/represent perceptual differences quantitatively(so that we know CIELAB is perceptually uniform)?
Staffjamesfong1
@estherc123 Great question. There are a couple of ways to do it. One simple experiment is to bring up two colors, and ask a test subject to judge whether or not they are different. 50% of the time, you give them literally the same color. The other 50% of the time, you give them two slightly different colors. The threshold beyond which a subject can detect the different colors with at least 75% accuracy can be thought of as a unit of color difference.
ncastaneda02
@jamesfong1 Color perceptions is a pretty disagreeable topic though and I would imagine that the variance is incredibly large for such an experiment even with a large sample size. Is there a similar test that is perhaps less prone to such human error? Or perhaps one that weighs samples based on how accurate the test subject generally is?
Staffyirenng
The way that the MacAdams study works conceptually is to ask human observers to compare two colors and say whether they can perceive a difference. A JND or "just noticeable difference" is the minimum difference where the human observer perceives the difference 50% of the time. The "MacAdams ellipses" drawn on the CIE diagram on the left are 10 JNDs in size.
akhilvemuri
Is perceptual uniformity usually desired for color schemes? Intuitively, I would think the frequencies spread evenly across all the colors, but maybe certain wavelengths correspond to more "color types," in which case uniformity might not allow one to represent all colors.
How do we measure/represent perceptual differences quantitatively(so that we know CIELAB is perceptually uniform)?
@estherc123 Great question. There are a couple of ways to do it. One simple experiment is to bring up two colors, and ask a test subject to judge whether or not they are different. 50% of the time, you give them literally the same color. The other 50% of the time, you give them two slightly different colors. The threshold beyond which a subject can detect the different colors with at least 75% accuracy can be thought of as a unit of color difference.
@jamesfong1 Color perceptions is a pretty disagreeable topic though and I would imagine that the variance is incredibly large for such an experiment even with a large sample size. Is there a similar test that is perhaps less prone to such human error? Or perhaps one that weighs samples based on how accurate the test subject generally is?
The way that the MacAdams study works conceptually is to ask human observers to compare two colors and say whether they can perceive a difference. A JND or "just noticeable difference" is the minimum difference where the human observer perceives the difference 50% of the time. The "MacAdams ellipses" drawn on the CIE diagram on the left are 10 JNDs in size.
Is perceptual uniformity usually desired for color schemes? Intuitively, I would think the frequencies spread evenly across all the colors, but maybe certain wavelengths correspond to more "color types," in which case uniformity might not allow one to represent all colors.