lipman@dtrc.dt.navy.mil (Lipman) (03/27/91)
Essentially we are looking for a perceptually equidistant rainbow looking color scale for a workstation monitor. We would specify the colors in HLS or RGB system. Please read comp.sys.apollo or comp.sys.sgi to see a longer version of this post. I forgot to post it over here. Thanks in advance, e-mail please Bob Lipman
maguire@cs.columbia.edu (Gerald Q. Maguire) (03/28/91)
See the paper in the SPIE proceedings (and several other places) about ~1980 by Steve Pizer of the Univ. of North Carolina on generating good color scales. He basically does a psychophysics experiment for each user and determine thes optimal color scale for each user + monitor combination. It turns out that this gives the highest amount of image transfer per pixel - as you have measured the just-noticeable color difference for the image chain in question. He has been doing these sorts of measurements for more than a decade - so you might want to contact him: smp@cs.unc.edu Chip
palmer@mcnc.org (Thomas C. Palmer) (03/29/91)
In article <MAGUIRE.91Mar27174136@cs.cs.columbia.edu>, maguire@cs.columbia.edu (Gerald Q. Maguire) writes: > > See the paper in the SPIE proceedings (and several other places) > about ~1980 by Steve Pizer of the Univ. of North Carolina on > generating good color scales. He basically does a psychophysics > experiment for each user and determine thes optimal color scale for > each user + monitor combination. It turns out that this gives the > highest amount of image transfer per pixel - as you have measured the > just-noticeable color difference for the image chain in question. > > He has been doing these sorts of measurements for more than a decade - > so you might want to contact him: smp@cs.unc.edu > Chip Has anyone tried generating colormaps by selecting colors along a line in a "perceptually linear" color space and converting back to RGB? Roy Hall's book ("Illumination and Color in Computer Generated Imagery") discusses two such spaces: L*a*b* and L*u*v*. Depending on just how perceptually linear these color spaces are, this seems like a reasonable approach. I've been meaning to try this but haven't found the time ... Perceptually equidistant colormaps should work fine for 2D image analysis applications. Has anyone even considered attempting to do the same thing for 3D stuff given that your colors are wildly transformed by the illumination model? The one saving grace is that the human perceptual system is pretty good at judging an object's color even under varying illumination conditions. -Tom "At 8,500 feet you rarely see a tax attorney or an obvious hairpiece." - Zippy Thomas C. Palmer North Carolina Supercomputing Center Cray Research, Inc. Phone: (919) 248-1117 PO Box 12889 Arpanet: palmer@ncsc.org 3021 Cornwallis Road Research Triangle Park, NC 27709