current page



Lecture on Nov 2, 2010. (Slides)

Guest lecture by Jason Chuang.


  • Required

    • Charting color from the eye of the beholder. Landa, Fairchild. (pdf, abstract in html)

    • Chapter 5: Color and Information, In Envisioning Information. Tufte.
  • Optional

    • ColorBrewer: Selecting good color schemes for maps. Cindy Brewer (html)

    • BruceLindbloom: Useful color information, studies, and Files. Bruce Lindbloom. (html)

    • Meet iCam: A Next-Generation Color Appearance Model (pdf, html)

    • Color2Gray: Salience-preserving color removal. Gooch, Olsen, Tumblin, Gooch. AMC Transactions on Graphics. (pdf, html)

    • A framework for transfer color based on the basic color categories. Chang, Saito, Naakajima. (pdf)


nchen11 wrote:

This is not related to color inasmuch that I wasn't sure where else to put it:

David McCandless: The beauty of data visualization

Another cool TED talk about data vis. :)

skairam wrote:

This also may be less related to color and perhaps more appropriate after the interaction lecture, but I wanted to share a link to the CAVE (computer-assisted virtual environment) project being used for visualizing data at the Idaho National Laboratory. It's pretty much a working Holodeck for data.

ankitak wrote:

I find use of color to be one of the most exciting parts of data visualization. However, one of the most important concerns that arise when we talk about using colors to encode data is its usefulness (or lack of it) for those who are colorblind. It would be interesting to study the color-perception of those who suffer from colorblindness and be able to design visualizations with color encodings that can be understood by everyone. In this vein, I found some interesting articles (in order of decreasing relevance):

sholbert wrote:

I really enjoyed the color lecture, especially the part on color perception. I thought the visualization of the survey of different cultures about what they call different colors was really cool.

Color is probably one of the easiest ways to make a visualization ugly and unreadable, so it is good to have some loosely structured protocols for using colors. Tufte gives some great uses of bad color in the reading, and the lecture really demonstrated how the combination of certain colors can result in erroneous visualization interpretations.

gneokleo wrote:

After reading the two readings I realized how tricky color can be. Encoding values with color hue or brightness can confuse some and choosing the right color is very important. On the other hand color can be very helpful and informative for the reader when done right. For our visualization for assignment 3 we decided to use color for some parts of our visualization and while we still have some tweaking to do in terms of the right colors I think it really helps the reader distinguish among sets of elements.

I also really liked how Tufte gives specific examples of bad use of color. It makes his points even stronger and I could immediately understand what he was saying.

msavva wrote:

@mattbush: I also felt that the part of the lecture linking color perception to language was extremely interesting and thought that many questions can be posed about the effects of social and linguistic backgrounds in this particular context. I was then reminded of a really nice book by Benjamin Lee Whorf, titled "Language, Thought, and Reality" which is all about how one's language can have a very big impact on thought patterns and perception as a whole.

msewak wrote:

This lecture on Color brought out a few properties of color that i hadn't thought about. I liked the Perceptual brightness part, the lecture highlighted how some colors may be perceived as brighter than others in a color palette. If used, they could be pre attentive, and draw attention to those fields encoded in those colors, even if the visualization did not intend for it to be so. How color is perceived is also affected by background, borders and lighting. Such things are important to consider while color encoding visualizations, and I hadn't thought of it this much. I liked Tufte's examples of bad use of color. The allocation of radio frequencies diagram really brings out ineffective use of color encoding as well.   

ankitak wrote:

This is the link to the video I had mentioned in the class - It is one of the talks from CS547 (HCI Seminar at Stanford) in 2009-2010. The title of the talk is Aesthetic Science of Color and it is delivered by Stephen Palmer from UC Berkeley. I found this lecture to provide a good insight in usage of color and its perception by different people.

mattbush wrote:

I have already seen some of this material in CS148, but I was pleasantly engaged with a lot of new material, such as the physics/frequency basis of color, the optical illusions, and the various measures and axes used for creating color spectra (physics-based, artistry-based).

I was hoping for more information on the cultural and emotional psychology of color, especially because that information is not covered in other CS classes, and because I believe it is just as relevant to making effective data visualizations just as much as what we studied today was.

Specifically: Earlier we studied the effect of cognitive biases on the perception of data visualizations, such as drawing attention to parts of our graph or making sure perception of numeric values are accurate. I think the emotional psychology of color would be a great way to implement drawing attention and giving priority to parts of visualizations.

rakasaka wrote:

I thought today's lecture was particularly interesting in light (no pun intended) of the fact that color is both quantitative and subjective (or perceptual). It's scary to think that there are so many colors to fit a massive Munssell book - I think knowledge encoding is good that we can only describe those colors with so many words!

I feel one of the important concepts is the translation of color from the online to offline world - the use of Munssell or Pantone color schemes may help in that regard, but visualizations need to be effective even when the correct colors cannot be employed. The idea of a color being "photocopyable" vs "printable" (colorbrewer) is perhaps the most salient representation of that fact. I think what also remains to be seen is how interactive visualizations can remain effective even when the interaction is removed or disabled.

jbastien wrote:

I have to link to the XKCD color survey, it's full of geeky neatness:

@msavva: the discussion on language determining which colors have names also reminded me of something I read recently "Does Language Influence Culture":

This very nice whiteboard animation on "The Secret Powers of Time" has an interesting section about languages and culture influencing perception of time:

yanzhudu wrote:

Color makes a huge difference in how people feel about a graph. A good color selection make a graph looks professional. However, choosing the right color is tricky. One suggestion is to used a image (for example, a natural scene) then extract color palette from it.

strazz wrote:

I think it's interesting how something that's inherently different to the perception of each person was structured in a way that allowed us to create standards for it. Both the readings and the lecture were very interesting as it allowed me to understand the physical process we go through when processing colors in any kind of activity, and this is something really important when designing applications or visualizations. However I was wondering about real world applications for all the research done on this area and I found this really interesting info-graphic that describes how colors affect purchases both in physical stores and online:

I found it to be very interesting since it combines elements from design, psychology, marketing and social context and sums it up on a graphic that makes a lot of sense.

estrat wrote:

Here's an interesting link that pertains to the discussion we had in class about what different people associate with different colors:

Not sure how scientific it is.

abhatta1 wrote:

Being from India, I would like to mention a few things which I have seen being used in Ayurveda [medicine of natural healing]. It associates different colors with emotions and feelings it evokes amongst people [for example red signifies angry] and the spots they affect the most. Some colors and their associations are red->survival instincts, orange-> strong pleasurable emotions, yellow-> sense of power, green -> calm,responsible, blue-> communication with the extended [sky,ocean,supernatural forces], indigo-> compassion etc. I personally think these are some of the points which might be taken into account while creating a graph. For example, putting too much red in the graph might evoke a feeling of angst in the viewer. There is a wide field of study called chromotherapy which deals with such things although I am not at all certain whether there is any scientific basis for it or not.

heddle wrote:

After this lecture, I would love to see more information on the break-down of colors by different groups of people. For examples, fuchsia is a very popular color lately, and for people who know a lot of colors do they give fuchsia it's own category? What about turquoise and aquamarine? @jbastien the xkcd link was fantastic for that. Now we need one that's designers versus engineers :)

The best part, though, was learning that there are really 4 primary colors. So basically, everything I learned in elementary school was a lie.

asindhu wrote:

I thought the optical effect with the picture of the castle was really amazing. Unfortunately I have to say I'm not completely sure I understand exactly why it happens, so if anyone wants to try explaining it that would be really helpful. I get the idea of opponent signals and how the brain has these three orthogonal responses -- brightness, red/blue, and blue/yellow, but I don't see how the effect we saw follows from that.

The biggest thing I got out of lecture today was that color is incredibly subjective, and it plays such a fundmental role in our daily lives that each person and each culture has a unique perspective on and understanding of colors. I thought the point about the risk in using uniform rainbow color scales was particularly compelling; because we group colors based on certain names, we perceive a bigger difference at certain points in the spectrum where a color boundary lies. I would assume this means that if we keep the color scale to a single color, like different shades of blue, then we minimize this problem?

emrosenf wrote:

@asindhu thanks for asking an awesome question. First off, here's the castle illusion, and here's how it was created. Explanation to follow.

amirg wrote:

I think color is one of the (many) topics of this class that goes beyond data visualization to any type of presentation in general (e.g., powerpoint presentations). Color seems to have a number of functions relevant to presentation in general: drawing attention to particular aspects, highlighting differences, enhancing readability, and perhaps even preventing boredom. Of course, color must be carefully used to serve all these functions.

On another unrelated note, I find color to be fascinating because their are major differences in the ability of people to perceive certain distinctions. I've always wondered about the conventions of using red and green for traffic lights, negative/positive changes, etc because of the prevalence of colorblindness. In particular, I think one problem is that these dichotomies of color have become cultural conventions, which makes it hard to break from them even if they are not optimal for the perception of many.

On yet another unrelated note, I was struck by the discussion of not using the rainbow spectrum to encode ordinality. While I agree that hues are not necessarily ordered, I think that in some cases conventions have taught us a particular ordering. For example, on heatmaps, I think that we have come to expect that red indicates hot while blue indicates cold and thus have learned a sort of hue ordering for this specific example. However, this obviously does not get past the problem of our tendency to group in our minds colors for which we give the same name, which causes us to form artificial boundaries in these types of representations.

emrosenf wrote:

@asindhu my understanding is that the cones in the opponent process model adapt to the stimulus (sensory adaptation). Notice that in the tutorial, to create the image, the image is inverted. What will eventually appear as a blue sky appears in the first image as 'orange', and what will eventually be green grass is 'blue'.

Because the cones adapt to the stimulus, when the stimulus is removed, there is a negative action potential. For instance, if "blue" has become the new baseline, then the absence of blue is "anti-blue". In RGB, if blue is (0,0,255), then wherever you saw blue before you now see (255,255,0) which is red and green. This is how you get the green grass.

Likewise, since orange and blue are opposite hues, the orange in the sky turns blue when the orange stimulus is removed.

Someone check my explanation. What am I missing?

trcarden wrote:

This reminds me of one of my favorite projects in the perceptual image space. Color sensitivity is actually surprisingly complex to model especially for just noticeable differences.

hyatt4 wrote:

@emrosenf and @asindhu I think I understand what is happening during the experiment from a conceptual standpoint - certain sensors being saturated. From a personal standpoint though I didn't get the effect. That is, when everyone in the room was saying, "Whoa!" I saw a black and white picture of a castle, which while nice, was not up there on the 'whoa' scale for me. Now I have some color vision deficiencies (I can't see the numbers in those colored spot tests) and I'm guessing that this would be just more proof of that to me that one my LMS sensors isn't firing quite the same as everyone else.

It seems like there would be someway to inverse filter a person's physiological color response. In the same way we can create audio filters in our stereo systems to make them sound like a band is playing in the Sistine Chapel (based on an impulse response taken from the Sistine Chapel). Then again, I suppose we would need to have some sort of contact with the person's retina to measure the response from white light to determine the response that would need inverse filtering - and that would probably be a bit messy. On second thought, I think I am okay with not seeing a colored castle for now . . .

jasonch wrote:

I've always found color theory very intriguing. In CS178, Prof. Marc Levoy did an excellent job on this topic (for 2 lectures). He also build some useful applets for understanding colors, gamuts, etc found here: If anyone is interested, the CS178 has good lecture slides on this!

andreaz wrote:

Color is such a fascinating topic because it taps into philosophical questions about human consciousness. It's interesting to think that no color actually exists in the world; it is purely the result of the color cones in our eyes and the neural circuitry in our brain. I also remember learning from a cognitive science class I took as an undergrad that our categorization of colors is different for different hemispheres of the brain; the effect that language affects color categorization is stronger when people view the color from their right visual field (compared to their left). The lateralization of language to the left side of the brain is said to explain this effect, as stimuli viewed from the right visual field is processed in the left side of the brain. The debate about linguistic relativity applied to color naming is also really fascinating ( and it's interesting to see Munsell's color system criticized by relativists for approaching color classification from a universalistic perspective.

Munsell's classification of colors according to the just noticeable difference reminds me of Stevens' work in the psychophysics of sensory functions. It's interesting that Munsell came from an artistic background, rather than a psychological one, and how that factored into his creation of a scheme that was meant for practical use, rather than something only descriptive. After learning about the Munsell color system, I now realize why it's so difficult to come up with a harmonic color scheme when picking randomly from using Photoshop's dropper tool, since the colors are never evenly spaced. It's easy to anyone to notice when a graphic's color scheme is slightly off, but after reading about Tufte's principles, and the color science research, it's great to be able to now pinpoint exactly which components contribute to a harmonic color scheme.

jdudley wrote:

I've built several websites for the health sector and through that I learned that green, blue, and orange are colors traditionally associated with health and wellbeing. When I first heard this, my BS detector went off, but as I started to look around, I saw the colors used sometimes exclusively at many health-oriented sites:

I still don't have an explanation for why these colors have an association with health. It could be that all these designers heard the same BS about the colors and followed suit, but that is unlikely to explain the full prevalence. My guess is that reds and yellows might evoke qualities of poor health (bleeding, jaundice?), but I'd love to get a better scientific handle on why the green/blue/orange color association is so prevalent in health and wellness.

iofir wrote:

I think that the lecture was very broad and introduced a lot of topics but did not really cover much depth. I think that the topic of color is very complex and important and it deserves more then a single lecture. color perception alone takes a good hour to cover deeply.

The one thing that I expected to hear about which unfortunately was not mentioned is color error/differences. or rather the different ways of measuring the differences (or error) between colors. He did mention how a unit difference in LAB color space should approximate the minimum noticeable difference. What was missing (in my opinion) was a short explanation about delta-E measurements and maybe a little about the differeces between dE94 vs dE2000.

Another pretty important concept that we didn't have time for was a good comparison between the gamete and accuracy of different color spaces. for example, colors like off-white and beige are very hard to represent in sRGB but a small (<0.25 dE) different is very noticeable to the human eye. That whole region of light colors is very hard to represent accurately in almost any color space.

Not that any of this is relevant to visualizations. it's very unlikely that you would use off-whites in your viz. just interesting stuff to know when talking about color science.

jsnation wrote:

I thought one of the more interesting things from the Tufte chapter was the discovery that colors from nature are generally more effective, due to their "widely accepted harmony". It also seems like when using color for quantitative information, it is best to have a varying value scale rather than varying hue.

I have a background in optical sciences, so we learned about the human visual system and color a fair bit. One thing that hasn't been mentioned really is that everyone's eyes are different, so they all have differing amounts of aberrations like chromatic aberration. This can come in the form of longitudinal or transverse. In longitudinal chromatic aberration, the error is that different color wavelengths converge at different distances within your eye, not all at the retina exactly, so that can lead to a blurriness in colors at either the high or low end of the visible spectrum, which in the extreme case could make points of one color look larger (and blurrier) than points of another. With transverse chromatic, it means that different wavelengths focus at different points on the retina, not at the same point. The good news is that in normal healthy eyes these aberrations are so small that you won't notice them unless you are specifically looking for them. But they do exist, so I would guess this might play into some of the variation between people's perceptions of color.

jtamayo wrote:

Matlab is used by a large number of scientists to build graphs to include in their publications. The default color map in Matlab goes from blue to red passing by cyan, yellow and orange. In many cases scientists won't change the default color-map of their heatmaps and 3D surface plots, we end up with a bunch of hard-to-read colormaps, especially for colorblind users.

This is a more reasonable colormap that people should use instead:

selassid wrote:

I think it's really cool that the absorbing molecule in the three types of cones in your eyes is the same. How does one cone absorb blue and another red then? The protein around the molecule differs and has a slightly different shape and charge distribution which alters the absorption spectrum of the molecule. Clever nature...

I also thought it was interesting that when you plot the observed color vs wavelength, the blues and purples seemed to bunch together relative to the greens and oranges; when Jason moved a box around the spectrum, it seemed to qualitatively encompass more colors at the long wavelength end of the spectrum. If you plot observed color vs frequency (energy), you seem to get a much more even distribution of the qualitative colors, though since wavelength is inversely proportional to energy. Perhaps frequency is a more relevant axis to view colors on when talking about perception.

mariasan wrote:

If you are interested in how to match color and what goes well together, there is a wonderful book by Shigenobu Kobayashi that's called "A Book of Colors". If you go to Amazon you can look inside to get a feel for it. I love this book, both for browsing colors and when I'm trying to pick color a schema.

ericruth wrote:

Like many others, I've studied the "scientific" aspects color in other CS classes, but this lecture was really interesting to me because of the neurological and perceptual aspects of color that we learned about. I think both of these aspects are as or more important than the scientific details when it comes to data visualization. However, one aspect I would have liked to learn more about is the artistic aspects of color. In particular, I'd be interested to learn how people develop good color schemes so we can experiment with our own color schemes with a bit more confidence. I realize that there are resources for color schemes (such as ColorBrewer and "A Book of Colors") - but it's a bit of a mystery to me how these color schemes are developed, and I think it'd be a really interesting thing to learn about.

wulabs wrote:

I think color is an important basic requirement needed in visualization. For instance if all of the colors were too similar (all shades of green) then the user would have a hard time distinguishing different parts of the graph. The idea of color in a visualization is similar to the idea of color in designing a web site. You just need a few (maybe 3-4) primary colors, and then the rest are secondary. After that, the actual color choices may make the visualization more pleasing but I highly doubt it has a significant effect on whether or not the visualization does its job

arievans wrote:

I've waited a little while to post a comment mainly because I wanted to let the concepts from the lecture kind of stir in my head a little bit. When I think back to everything in the presentation, what really sticks out in my mind is the demo with the castle and the colors. I'm really interested in diving deeper into learning how that works.

But more importantly, I think there are applications for that technology beyond just a cool science experiment. The only potential issue with it is that you have to be staring or have the user's attention for a while so that the desired effect can take place. What I'm wondering is if somehow we can achieve that same effect subliminally--i.e. we set up the coloration in the background and obstruct it make a character (or something else) occupy the viewers attention in such a way that where they are staring is close to where the black dot would be. Then, we could move the character away and remove the coloring and have the background be colored. I suppose that in the film world you dont really need to do that since you can just achieve that effect through digital manipulation, but I think it's difficult to mirror how this effect works on an individual basis. Using this technique would allow for multiple variations for the viewer and perhaps new interpretations on the same view depending on how the colors end up making the view come to life.

Just throwing out an idea--feel free to experiment with it or let me know what feedback you have! Excellent presentation; Very interesting!

esegel wrote:

My favorite example from the readings is on page 94 of Tufte's "Envisioning Information" showing a graphic with grey squares on a grey field, and showing how the grey figures clearly separate from the ground with the tiniest adjustment in border color. It is a small subtle example that hammers home most of Tufte's design principles and general aesthetic. First, that bold lines and color should be used sparingly. And second, that the same effects can be achieved with slight design variations. His version of "Less is More" and "Chart minimalism". As a result, the focus of the graphic is on the content/data itself, rather than the features of the design.

adh15 wrote:

@esegel, I also liked the example showing the powerful effect of a thin border. On a similar note, I was also intrigued by the use of the thin blue line to shift the perception of the red ink on the road atlas.

I am also interested to hear accounts from people who were "taught a systematic method of color specification at an early age" as noted in the Landa and Fairchild article. I wonder what effect such education has on later abilities involving color (art, design, etc.).

nikil wrote:

I enjoyed the biological basis of color in the discussion. I was surprised to hear about all of the differences in different people's perception of colors, as I had just assumed that pretty much everyone saw color the same way (besides disabilities); =>I definitely think about people's perception in visualizations.

sklesser wrote:

Color is a really fascinating subject and using different color spaces has real advantages and isn't restricted to a background technical detail. In Photoshop generating good masks and selections of a difficult to define or fuzzy object is a difficult task. One approach is to use a particular channel such as the red channel or blue channel which approximates the selection and then adjust levels to get a good edge and then fill in the inside and outside. However this technique relies on a clear color transition between foreground and background objects which isn't always present.

An alternative I find useful is to convert the image into LAB color space which gives you a luminosity channel and two color channels as opposed to three color channels. You can then use the same technique described above with the luminosity channel and extract objects from the foreground which have a different brightness than the background. This is especially useful for pictures taken with a flash since a flash lights up foreground objects far more than background objects.

Leave a comment