> On 12/5/06, *Ged Ridgway* <[log in to unmask]
> Personally, I don't like these exciting all-around-the-spectrum
> colormaps (which have slightly blueish-green and slightly
> reddish-yellow looking kind of similar(ish), but on different sides of
> zero). I would prefer a simpler straight line through the colour
> wheel, like cyan-black-red, but with so many list subscribers studying
> perception/vision/psychology I wouldn't dare to claim that this would
> look perceptually better...
In case anyone is curious, the image at
http://www.cs.ucl.ac.uk/staff/G.Ridgway/vbm/cmaps.png
shows a "testcard" with equal increments of intensity (columnwise)
over the full range, with the three colormaps:
* jet - Matlab's default
* actc - from the [log in to unmask] presumably created by
Matthew Brett, which can be easily adapted to be a generally usable
Matlab colormap, as here:
http://www.cs.ucl.ac.uk/staff/G.Ridgway/vbm/actc.m
* cyanred - a quick implementation of my randomly proposed
cyan-black-reg, which the keen can find here:
http://www.cs.ucl.ac.uk/staff/G.Ridgway/vbm/cyanred.m
As I said, I don't really want to make excessively bold claims, but I
think my simpler colormap:
* shows more perceptually linear increments (especially the larger
steps along the rows, where jet and actc show some quite marked
non-linearities, at least with my monitor/eye/mind)
* makes clearer the small increments down the columns (especially at
the extremes, e.g. down columns 2 and 9 seem particularly bad in jet/actc)
* is more perceptually symmetric about the middle values
* and, if used for symmetric positive/negative data, makes it much
clearer whether a pixel is positive or negative than a kind of
yellowy-bluey-greenish does.
But hey, it's a free world, no-one has to agree with me...
Goodnight,
Ged.
|