(1)
– Colorbrewer produces color schemes that have few visual artifacts, while maintaining visual appeal.
– Colorbrewer varies all three of hue, lightness, and saturation even for one dimensional (diverging, sequential) data.
– I’m curious about the use of light beige (255, 255, 204) in several of the qualitative schemes: it looks very close to white and the white background used for the map.

(3)
This article recommends against the use of rainbow color map in visualizations: rainbow maps are
– not iso-luminant
– cannot be easily ordered
– has bad spacial resolution
– contains banding artifacts

I wonder if the following `fixes’ will work:
– use luminance in addition to color, avoiding low spacial resolution
– use hue from cylindrical h*uv o achieve iso-luminance and remove banding artifacts

I was disappointed with Figure 1b, which is intended to demonstrate that rainbow colors do not have a obvious ordering that can be naturally visualized by a human.
Figure 1b contains two sequences of four circles each, with colors [R Y G B] and [R G Y B].
However, this figure has two major deficiencies.
– [R Y G B] and its permutations are hardly the rainbow colors; the rainbow contains more than four colors, and six or seven distinct bands.
– With a coin toss, the ordering of [R Y G B] is correct 50% of the time, which, depending on the baseline, can be considered high.

(4)
Hue preserving color blending:
This article presents a very nice an simple method to add colors without adding inadvertent hues. I like the concrete perceptual basis for the algorithm, but I thought the paper was too drawn out in explaining the algorithm.

(5)
The very large set of color usage examples is valuable as reference for when one doesn’t know how to use a particular color.

The rainbow color map topic came once before during the earlier part of the readings. I think Munzer’s reading raised this issue of avoiding the rainbow maps. But this reading also suggested different alternatives. This part was actually helpful and will help us decide appropriate coloring for our data. The author also give some statistics from IEEE visualization conference papers related to number of visualization still using the . The goto statement anology is interesting, but, I think rainbow map might be more dangerous than goto statements. Its probably still fine to use goto statements occasionally.

The two main points that I got from these readings (at-least from rainbow paper) are (i) don’t use rainbow map (ii) look at Colorbrewer when picking colors.

In Rainbow article, black-body radiation was new to me. I couldn’t understand what contours filled with a constant color is. It would be good it has the example.

Designer’s color blog was very interesting and I can see the big effect of subtle changes in color.

Colorbrewer would be very useful when choosing colors.

Ware’s writing goes on to reinforce this principle, noting that the luminance channel is not only perceptually important with respect to color, but also with respect to a variety of other perceptual phenomena. In addition to the discussion of the luminance channel, the paper goes on to focus heavily on how contrast effects can skew the user’s perception of color. In general, this work seemed to provide a scientific backing to Brewer’s conclusion in one that dimensions in the color mappings should correspond to a logical ordering of the data.

However, one interesting note is that Ware mentions the monitor spectrum as being a preferable color ramp. However, the third paper by Rhyne essentially trashed the idea of using full-spectrum ramps by asserting that rainbow color ramps. She argues that such ramps lack a perceptual ordering, have uncontrolled luminance variation, and introduce data-independent gradients, all points which Ware also addresses as being inappropriate for color ramps.

Ware’s paper also mentions that saturation is a bad choice for color ramp creation as it varies in the yellow-blue and red-green opposing color channels. However, paper four “Hue-Preserving Color Blending” does exactly that to provide a color-blending scheme that the author asserts is an improvement over traditional color ramps. This technique, however, does address Rhyne’s comments on rainbow ramps introducing false divisions by sharp variations in hue by only converting hue over a gray color space. However, this same grey color space can cause issue because it will be present across all ramps utilizing this sort of blending.

Rhyne’s comments on black-body radiation blending brought in a new, biological element to color. Essentially, the black-body model is an expression of encoding heat by color. This ramp may be interesting as an effective choice from an evolutionary standpoint (as in heat implying warning), but this point is not explored in any of the reading. However, this idea of psychologically and culturally-based color ramps is echoed in the fifth reading, which is somewhat interesting as the work is written from an entirely emotional and non-scientific standpoint.

For the “Rainbow Color Map Are Harmful” paper, I agree the authors’ point that the rainbow color map is not suitable for continuous data. However, I thought the prevailing usage of rainbow color map is not as unreasonable as the authors indicated. For example, in my own field, a lot of simulation softwares using rainbow color maps for their results (e.g. ANSYS(R)). It doesn’t cause many trouble because in our engineering field, most time we are more care about the ranges of certain data, or the peak and lowest value within a certain data set. If we want to see the exact value of a data, in my opinion there would be no methods better than a scatter (and/or curve) plot. So, I think the harmness of rainbow color map is actually determined by the application that the maps are used in. The authors of this paper might be exaggerating the damage of rainbow mapped color.

ColorBrewer is an useful tool in our domain. It’s quite object oriented since for each kind of data there are different schemes to fit in. For example, “diverging” schemes looks very similiar to a Smith Chart in the Microwave field and might become a substitution.
The “Hue preserving color blending” paper is also a good paper and the results listed by so many figures are really convincing. However, I’m a little bit confused by the application of new approach: why and how the approach is good for perception? For example, in Fig 9 (b), the yellow hue is eliminated, but it makes the audience harder to distinguish from the brain and the body.

The “Rainbow Color Map Are Harmful” paper was intriguing. I didn’t think before how bad that color scheme is. On the contrary I had even often thought about using that on many occasions because it has so MANY colors available. Luckily, the reading was followed by Brewer’s “color brewer”. Her approach to visual typology for various data types look very effective, and unless there is a semantic need to designeate specific colors I can easily picture myself revisiting the engine over and over. Naturally, the next reading (The Meaning of Color) was on the semantics of color. It would be great if there is an additional reading that explains the underlying mechanism behind the meanings described in that article.

The “Hue preserving color blending” paper explains how to make non-succesive components stand out in blended colors, by preserving hues. Together with Ware’s “Color Sequences for Univariate Maps” paper which is an elaboration on his arguments on luminance in the book chapter last week, I think it gives us a nice set of color choices to use for making significant patterns more visible in data visualizations.

The paper about Rainbow Color Map by Borland and Taylor emphasizes that these kinds of maps have bad feature comparing to alternative choices, in illustrating data. What I learned in this paper is the choice for different data, like nominal, high-frequency ordinal, color map and interval and ratio data. Among all these popular data, there are always proper plots to characterize them.

ColorBrewer is a fancy tool. There are color schemes for different data, including sequential, divergent and qualitative ones. “Sequential” schemes can be applied to quantities such as population and height, and “diverging” schemes, which contain a central point from which quantities may be increasing or decreasing.

The paper, A Perception-based Color Space for Illumination-invariant Image Processing, introduces a new color space. The theoretical foundation here is a new metric, perceptual metric, which is converted from the tristimulus values. I do not quite understand the application of this method.

‘Color theory for designer’ maps colors to certain characters, which represents what people think about those colors. The main goal is for website design and a right color selection may reflect the theme precisely.

I never expected that this Dijkstra style paper will change my knowledge about colopmap that we have been taking for granted. With the beautiful colors of rainbow everyone was pleased, so it never occurred that we were telling lies (or hiding information) with it. May be with a side legend, we were able to compare the results and never thought that perceptual ordering was also an important issue.

I will try “Black-Body Radiation Map” in future work and see how I can convince myself that a new color map is better ( and natural ) than old rainbow map.

Color Sequence for Uni-variate Map;
******************************************

This was more or less a repetition of the last reading on colors. This paper also reinforces that luminance is better for shape identification and colors are good metric information.

Looking at the figure #2, I think that I agree with first and the last color map results, but not fully understood second, third and forth. I was wondering if we have to use only two colors, then why not choose black and white only ( first color map ).

The final summary of the paper is good one and must be taken into full consideration all the time.

Paper : A Perception-based Color Space for illumination-Invariant Image Processing
**************************************************************************************************

A really wonderful paper with some practical applications. Although I couldn’t understand its mathematical formulations, but in my view, this technique must be part of core modules in all the commercial software packages dealing with segmentation and matching.

Hue-Preserving Color Blending
**************************************
This paper presents a practical problem and provides a very simple and an elegant solution. The algorithm presented in the paper is straightforward and the results look really impressive. Henceforth, I shall apply this technique in all my work.

I found Poynton’s FAQ to be an absolutely brilliant resource that I will go back to many, many times until, eventually, I understand and am able to commit all of it to my brain. I wish I had seen this *before* last week’s class on color.

Borland’s appeal to ditch the rainbow color map is a very useful read. I did not realize that the rainbow scheme was so prevalent in spite of it being ill-suited to conveying sensible information (but then, I look at Internet Explorer or MS-Windows and realize that bad design doesn’t lose, but in fact, can thrive).

I have used Cindy Brewer’s colorbrewer application in my work, and find it to be a very useful device.

Smashing Magazine’s article was fun and easy, illustrating its point through web graphics, matching moods and temperaments with color schemes.