View Issue Details Jump to Notes ] Print ]
IDProjectCategoryView StatusDate SubmittedLast Update
0002459Old ParaView Bugs public2005-11-07 16:062005-11-07 16:08
ReporterAmy Squillacote 
Assigned ToBerk Geveci 
PrioritynormalSeverityminorReproducibilityalways
StatusassignedResolutionopen 
PlatformOSOS Version
Summary0002459: changing the predefined color scales
DescriptionThis is from an e-mail from Russell Taylor (my former advisor at UNC).

On my continuing mission entitled "The Rainbow Color Scale Must Die!" I'm happy to provide you with some ammunition for how to change it. The grayscale and CIE blue/gray/red in ParaView are good ones and should remain. I attach a number of color scales that would be more useful. Let me know if you have trouble interpreting the files. You might also consider one that truncates the black-body (blk-red-yellow-wht) at red, so that there is no black in it -- this will be better for mapping onto surfaces. You'll also want an isoluminant grey-to-red scale for cases without a special "zero."

Note that the only rainbow we have is rainbow-banded. If included, it should be modified to remove two of the three very-similar greens and perhaps one of the two similar blues.

The "standard" rainbow should be replaced by the blackbody, and the rainbow renamed "Suboptimal" in VTK and Paraview and anything else that ya'll do. I understand that complete removal would break all the tests, which seems really bad.

This removal is VERY IMPORTANT. Colin Ware describes why in "Information Visualization: Perception for Design." I summarize this and the reasons for choices in one of my vis course lectures (http://www.cs.unc.edu/~taylorr/Comp215/slides/2005_02_17_2D.pdf [^]). I'm willing to argue the issue up the chain of command as much as it takes. The best thing would be to invite Colin up for a day-long workshop for Kitware employees on perception, color choice, 3D shape, and so forth. I have "canned" slides that he could use for such (from my Vis course) if it helps to entice him.

The basic argument: The rainbow color scale is never optimal (note, the banded one can be). (1) It has large flat areas with no apparent change in hue separated by sudden changes. This introduces FALSE structure into any data set mapped to it. (2) It is isoluminant, so is not able to display high-frequency information (like what happens when you view green text on a red background). Note that this is a bonus if you want to avoid distorting surface shape, but there are other color maps that do this better. (3) It is not perceptually ordered: you have to either remember "ROY G. BIV" or else look back and forth to a key to remember which color is "more" than another.
TagsNo tags attached.
Attached Files? file icon rainbow-banded [^] (545 bytes) 1969-12-31 19:00 [Show Content]
? file icon blk-red-yellow-wht [^] (151 bytes) 1969-12-31 19:00 [Show Content]
? file icon purple-gray-cyan [^] (156 bytes) 1969-12-31 19:00 [Show Content]
? file icon blk-green-cyan-wht [^] (151 bytes) 1969-12-31 19:00 [Show Content]
? file icon blk-golden-wht [^] (150 bytes) 1969-12-31 19:00 [Show Content]

 Relationships

  Notes
There are no notes attached to this issue.

 Issue History
Date Modified Username Field Change


Copyright © 2000 - 2018 MantisBT Team