From CS 294-10 Visualization Sp10

Jump to: navigation, search

Lecture on March 10, 2010




  • Color and information, In Envisioning Information, E. Tufte
  • A rule-based system for assisting color map selection, Bergman, Treinish, Rogowitz, (html)
  • Color guidelines, Brewer, (html)


Jon Barron - Mar 09, 2010 04:25:47 pm

Tufte: I found it difficult to draw coherent rules from this chapter. One lesson seems to be (in typical Tufte fashion) to be minimalist when using color. Another: don't use rainbow color schemes.

IBM: There are some good lessons to be learned here. I particularly like this: "saturation-based colormaps, which display variations in the magnitude of a hue, would be inadequate for conveying high spatial frequency information, but well-suited for representing larger-scale spatial variations." The paper did not make it very clear how they're using this philosophy. I understand the FFT method that they compare their rule-based system to, which is not very helpful.

For the most part, I have no idea what this application they're describing does. Sadly, things would be much clearer if this paper had more visualizations --- or better yet, a video. I really like the colormap construction tool, mostly because I feel like I kind of understand it.

Brewer: This is really cool! Things I like: The separation between qualitative, diverging, and sequential color schemes, and the little icons that indicate properties of the color scheme (especially the fact that they indicate colorblindness, a topic that seems to have been overlooked in the other papers).

Jeffrey Patzer - Mar 10, 2010 01:01:48 am

As I was reading Tufte's chapter, there was a particular sentence that I didn't really understand: "Ideas not only guide work, but also help defend our designs (by providing reasons for choices) against arbitrary taste preferences". While the rest of the chapter provides excellent guidelines on how to effectively use color when encoding signals, this particular phrase makes no sense. While I can agree that Ideas do guide one's work, I don't know if they are the defense of the design. Rather the data behind the idea is the defense. I just found this statement to be completely arbitrary. However, his ideas on how to capture color, and how to layer it do make a good point. The use of nature as a guide is awesome. Nature really does use mostly muted colors except for especially bright objects that need to stand out from everything else.

Arpad Kovacs - Mar 10, 2010 12:53:37 am

Tufte seems to contradict himself when discussing the accuracy of color perception. First, he states that "color is a natural quantifier, with a perceptually continuous (in value and saturation) span of incredible fineness of distinction, at a precision comparable to most measurement." Yet later on, he argues that color to quantitative translations are "nonlinear (thus gamma curves), often noisy and idiosyncratic." He seems to be trying to say that since there is such a tremendous range of colors, it is excellent for nominal measurements. Unfortunately, due the high variance and lack of calibration in the human visual system, the ultimately color should not be used for standardized quantitative measurements. Alternatively, accurate, color-based quantitative readings are possible if the information must be redundantly encoded, either through contour lines as in the ocean depth map, or numerical labels like in the Pythagorean theorem proof example.

Tufte's advice for using color is quite straightforward:

  • use it sparingly, so that things you want to highlight do stick out due to contrasting color.
  • be careful about interaction/interference; different contexts can make identical colors appear different.
  • use nature as inspiration for your pallete (hence, avoid those outrageous rainbow color schemes).

In summary, color can be extremely effective for attracting attention to a particular point of interest, or making very rough estimates / nominal comparisons. One thing Tufte did not mention is that when some colors are used in combination, we tend to associate them with particular connotations. For example, red/green are generally perceived to mean stop/go, while blue/red mean cold/hot. I am curious whether these cultural connotations subconsciously impact our perception of color, and if so, what we could do to minimize this interference.

DavidZats - Mar 10, 2010 03:10:45 pm

The three papers in today's reading discuss how color should be used to effectively present information in visualizations. The chapter from Tufte's book discusses ideas such as the using bright colors sparingly to avoid jarring the reader. This chapter also shows how the use of very thin lines in addition to colors can more effectively convey differences to the person viewing the image. As shown in the chapter, think lines are largely unnecessary and can be somewhat jarring.

The paper describing a new Colormap tool attempts to assist users on picking mappings that will accurately inform the viewer. This tool takes both statistical information about the data and the motivations of the user into account when suggesting a set of usable Colormaps. Since this tool seems to be so effective, why has it not been incorporated into commercial projects?

Finally, the last paper provides basic advice on when hues, saturation, and luminance should be used to effectively convey information.

Zev Winkelman - Mar 11, 2010 05:12:17 pm

I found the guidelines of the Brewer paper to represent what I am enjoying most about this course: well thought out, generally applicable rules, for design choices that we are confronted with all the time.

In the past, when confronted with a choice of what kind of color scheme to use, I relied on an my own intuitive, yet unstructured process to arrive at a suitable solution.

Sometimes this involved trial and error, sometimes it was reuse of things that had worked in the past. But rarely did it involve a systematic evaluation of the characteristics of the data and consideration of what kind of color scheme would best fit.

Yet another tool tucked in to the repertoire :)

Jonathan Yen - Mar 11, 2010 05:46:38 pm

I found these readings to be pretty "colorful" in demonstrating different uses of colors. The taxonomy of colormaps in the IBM paper was pretty neat and gave a good sense of how to approach the use of colors for the nature of the task. Brewer also does a pretty good job of explaining the different types of color schemes and how to approach them, but it would have been nice to have some visual examples in her paper. I think it would be nice to know also about the cultural implications of colors and how different colors can express different moods.

Stephen Chu - Mar 13, 2010 05:17:10 pm

I will remember Tufte's suggestion of using colors found in nature (except rainbows). I don't know why I never realized this strategy when struggling to choose colors for visualizations, but it is so intuitive to use colors that feel natural to the human eye. I would have liked to see more examples of Tufte's favorite and most-used color schemes. And because a significant portion of the population is color-blind, it would have been interesting to read Tufte's thoughts on how to deal with this problem.

Boaz Avital - Mar 17, 2010 05:36:30 am

I liked the idea of verifying that your color graphic looks good in color by checking if it still looks good in grayscale (luminosity only). I'm very fond of these kinds of tools or quick tips/tricks to making better visualizations.

When talking about color, Tufte seemed to translate in his other ideas. The gist was basically an ink to information argument, where instead of ink you measure color and vibrance of color. Sparser is better, the Tufte way. He also brought it again the idea from layering and gridlines that the background should be as muted as possible. This suggestion is obvious but is such a help to explicitly specify. It's something that a lot of people could have learned from, especially in, say, the early days of web design.

Ebby Amirebrahimi - Mar 17, 2010 05:54:26 am

I thought Brewer's article on color was very useful. Particularly, I found his breakdown of color schemes to be helpful. There are three types of color schemes. The first are Sequential Schemes which can be used to show a gradient of intensity, such as low to high. The next are Qualitative and Binary Schemes which are basically used as a key to represent different elements. Next, Diverging Schemes can be used to both show different elements as well as emphasize important data of those elements. An example given is of a rainbow scheme used for scientific visualizations. Finally, two-variable schemes can be created to implement various schemes on different axes for example.

RyanGreenberg - Mar 17, 2010 12:29:34 pm

I found Tufte's suggestion of using colors "found in nature" interesting. Although it doesn't guarantee that you'll follow the other suggestions Tufte presents in the chapter (obviously you can find garish tones and rainbow hues in nature as well), this could be a neat way to generate a color palette. If it doesn't already exist, an interesting tool would be to select discriminable colors from a provided sample photo, which allows a user to define a color swatch by example.

Brewer's paper had many more practical suggestions, including this great aside on aesthetics:

"I recommend that you do not become overly concerned about which colors your audience will like. Everyone seems to have an opinion about color aesthetics, and members of your audience will undoubtedly have differing opinions based on their own color preferences. [...] focus your attention on organizing the perceptual dimensions of color so that your data is presented clearly, whether or not you’ve picked everyone’s favorite colors."

I read this paper a few months ago which suggests using a red-yellow-blue model in some situations over RGB because people have better intuitions about mixing colors in subtractive models. I'm not sure if this is the case, and I didn't see anything in these readings that supports that particular point.

Chetan Nandakumar- Mar 17, 2010 04:38:11 pm

I'm particularly interested by the combination of colors on attention. It seems that different compositions or combinations of color would draw the attention more than others.

If this is true, it seems that it would be a very important guide for visualization as those color conjunctions would be more salient to the observer.

Priyanka Reddy - Mar 18, 2010 06:09:53 am

I've been playing with the Color Brewer tool and Color Brewer 2, and I think it's a fun tool to try out and explore. I could see this being useful to me. There were a couple features I really liked. First, the legend type was nice to have. Not only does it give you the colors that are most appropriate for your use case, but for those who aren't aware of the different legend types can really learn from this. The second feature was the various icons on whether the color scheme was color blind friendly, photocopier friendly, etc. This is a feature I've never seen anywhere else, and it's such a familiar problem that I'm happy to see it included here.

I found the discussion on color usage in lecture to be compelling, especially the idea that you can use colors that are obviously distinguishable from each other, but because they're not standard colors, they become really hard to talk about. This involves really thinking about the use case of the visualization and the user experience that surrounds it, not just the visualization alone. I think understanding more about how users react to and use visualizations will make for better visualizations.

Akshay Kannan - Mar 18, 2010 12:19:16 pm

Tufte's views on color basically seem to be to avoid any colors that are too jarring and to stick with a natural color palette. In an earlier chapter, I recall Tufte's discussion on how color can also be used to establish contrast- for example, data labels or grids could be presented in a separate color from the rest of the visualization to allow them to stand out and prevent them from interfering with the other parts of the visualization.

I found Brewer's paper on color usage to be interesting, especially in discussing the distinction between RGB and CMY, as well as the discussions of various color schemes.

Mason Smith - Mar 24, 2010 11:34:45 pm

Tufte: I really like the use of color with Euclid's Elements. I'm somewhat skeptical of how well it scales, but the examples demonstrated in the book were fairly sizable. I think the most useful non-obvious advice (given everything else we've seen up to this point in class) was advocating the use of "natural" colors in images.

Shimul Sachdeva - Mar 24, 2010 8:01:35 pm

While some PRAVDAColor applications are interesting in general, its usage for 3D applications is especially nice. For instance, the 3D map showing ocean conditions used hue and luminance to show 2 variables together in a fairly easy and readable format. It was interesting to note that luminance component conveys monotonicity in the data for high spatial frequency data, while the saturation component can be used to convey monotonicity in low-spatial-frequency data. I had not heard of CMY and HVC color systems until I read the ASA paper; although the systems serve different purposes as the RGB system, there are very few applications I have come across that use CMY or HVC. Since CMY is the system that printers follow, it seems logical to have more graphics applications use CMY instead of the traditional RGB system.

From the lecture on color, I especially liked the example, Bezold effect and the Munsell Atlas. The discussion on simultaneous contrast was also interesting. I also like the bulleted slide on color design principles.

Prahalika Reddy - Mar 29, 2010 10:41:53 am

The lecture on Color was actually quite interesting. I liked the information on the 2D Color Spaces, although I don't understand it fully. What determines which colors are in each of the different spaces (Normal, Protanope, Deuteranope, and Tritanope)? Also interesting was the Color Appearance slides. I think it's pretty cool how the same color looks so different when the background is different or the lighting is different. I agree that choosing the right colors is a very important design decision, and choosing the wrong ones can make an entire visualization less effective. It's especially important in complex visualizations, such as maps.

Jaeyoung Choi - Apr 02, 2010 09:23:09 pm

Brewer - This reading is really helpful in understanding how to use color scheme to represent data visually. It states the rule of thumb of how to use three perceptual dimensions (hue, lightness, saturation) of color separately or combined. I was surprised with the high percentage of population having color vision impairment and I found the part where diverging scheme is used with spectral (or rainbow) schemes to represent the data clearly even to those who are color blind to be quite interesting.

[add comment]
Personal tools