Dear r-devel,
I have implemented a function to create color palettes with improved
contrast and logical order, compared with the built-in ones:
source("arni.colors.R") # code is given below
barplot(rep(1,100), col=arni.colors(100), space=0, border=0, axes=F)
par(mfrow=c(2,1)) # rainbow() has too much green...
barplot(rep(1,50), col=rev(rainbow(50,end=0.7)), space=0, border=0,
axes=F)
barplot(rep(1,50), col=arni.colors(50), space=0, border=0, axes=F)
arni.colors(100, plot=T) # graph color recipes
arni.colors(10, rgb=T) # numeric color recipes
The arni.colors function also includes a palette="depth" which is an even
progression from dark blue to light blue. The color mixing proved to be
more complex than I anticipated:
arni.colors(100, palette="depth", plot=T)
I'm hoping that the R Core Team can adopt these palettes, as is, as
separate functions, or as arguments to the already existing palette()
function. Of course I don't mind if the R Core Team changes the function
name or any other part of the code.
Thanks,
Arni
arni.colors <- function(n, palette=c("temperature", "depth"),
rgb.matrix=FALSE, plot.colors=FALSE)
{
rgb2hsv <- function(v) rgb(v[1], v[2], v[3])
palette <- match.arg(palette)
x <- seq(0, 1, length=n)
if(palette == "temperature") {
r <- 1 / (1+exp(20-35*x))
g <- pmin(pmax(0,-0.8+6*x-5*x^2), 1)
b <- dnorm(x,0.25,0.15) / max(dnorm(x,0.25,0.15))
}
else {
r <- 0.6*x + 0.4*x^2
g <- 1.5*x - 0.5*x^2
b <- 0.36 + 2.4*x - 2.0*x^2
b[x>0.4] <- 1
}
rgb.m <- matrix(c(r,g,b), ncol=3,
dimnames=list(as.character(seq(length=n)),
c("red","green","blue")))
hsv.v <- apply(rgb.m, 1, rgb2hsv)
if(rgb.matrix)
attr(hsv.v, "rgb.matrix") <- rgb.m
if(plot.colors) {
opar <- par("fig", "plt")
par(fig=c(0,1,0,0.7), plt=c(0.15,0.9,0.2,0.95))
plot(NA, xlim=c(-0.01,1.01), ylim=c(-0.01,1.01), xlab="Spectrum",
ylab="", xaxs="i", yaxs="i", axes=FALSE)
title(ylab="Value", mgp=c(3.5,0,0))
matlines(x, rgb.m, col=colnames(rgb.m), lty=1, lwd=3)
matpoints(x, rgb.m, col=colnames(rgb.m), pch=16)
axis(1, at=0:1)
axis(2, at=0:1, las=1)
par(fig=c(0,1,0.75,0.9), plt=c(0.08,0.97,0,1), new=TRUE)
midpoints <- barplot(rep(1,n), col=hsv.v, border=FALSE, space=FALSE,
axes=FALSE)
axis(1, at=midpoints, labels=1:n, lty=0, cex.axis=0.6)
par(opar)
}
return(hsv.v)
}
______________________________________________
[EMAIL PROTECTED] mailing list
https://www.stat.math.ethz.ch/mailman/listinfo/r-devel
