Some useful color tables for images and tools to handle them.

Several color scales useful for image plots: a pleasing rainbow style color table patterned after that used in Matlab by Tim Hoar and also some simple color interpolation schemes between two or more colors. There is also a function that converts between colors and a real valued vector.

Usage

tim.colors(n = 64, alpha=1.0)

larry.colors()

snow.colors(n=256, alpha=1)

data(Mines.colors)

two.colors(n=256, start="darkgreen", end="red", middle="white",
alpha=1.0)

designer.colors( n=256, col= c("darkgreen", "white", "darkred"), x=
                              seq(0,1,, length(col)) ,alpha=1.0)

color.scale(z, col = tim.colors, NC = 256, zlim = NULL,
                 transparent.color = "white", eps = 1e-08)

fieldsPlotColors( col,...)

Arguments

alpha: The transparency of the color – 1.0 is opaque and 0 is transparent. This is useful for overlays of color and still being able to view the graphics that is covered.
n: Number of color levels. The setting n=64 is the orignal definition.
start: Starting color for lowest values in color scale
end: Ending color.
middle: Color scale passes through this color at halfway
col: A list of colors (names or hex values) to interpolate. But for the color.scale function this can be also be a function that returns NC colors when called with just the NC argument.
x: Positions of colors on a [0,1] scale. Default is to assume that the x values are equally spacesd from 0 to 1.
z: Real vector to encode in a color table.
zlim: Range to use for color scale. Default is the range(z) inflated by 1- eps and 1+eps.
transparent.color: Color value to use for NA's or values outside zlim
eps: A small inflation of the range to avoid the boundary values of z being coded as NAs.
NC: The number of colors to return from calling the function passed in the col argument. This is only used if col is a function.
...: Additional plotting arguments that are passed to image.plot.

Details

The color in R can be represented as three vectors in RGB coordinates and these coordinates are interpolated separately using a cubic spline to give color values that intermediate to the specified colors.

Ask Tim Hoar about tim.colors! As a Mattlab black belt this was his favorite scale at NCAR in that system. The viridis package has a similar version (turbo) that is a grey scale friendly alternative.

two.colors is really about three different colors. For other colors try fields.color.picker to view possible choices. start="darkgreen", end="azure4" are the options used to get a nice color scale for rendering aerial photos of ski trails. (See https://github.com/dnychka/MJProject.)

larry.colors is a 13 color palette used by Larry McDaniel (retired software engineer from NCAR) and is particularly useful for visualizing fields of climate variables.

snow.colors is the scale used by Will Klieber's team for visualizing snow cover from remotely sensed data products. See the commented code for the script as to how how this was formed from an orignal raw 256 level scale. Note the that first color in this table is grey and is desigend to represent the minimum value of the range ( e.g. 0). If the image is in percent snow cover then zlim=c(0,100) would make sense as a range to fit grey pixels to zero and white to 100 percent.

See data(Mines.colors) to define the data sets that are web colors for Colorado School of Mines. Note that since these are defined as text strings do _not_ quote them when using in the R graphics functions – they are already character data. E.g. plot(x,y, col=blasterBlue)

designer.color is the master function for the other scales. It can be useful if one wants to customize the color table to match quantiles of a distribution. e.g. if the median of the data is at .3 with respect to the range then set x equal to c(0,.3,1) and specify two colors to provide a transtion that matches the median value. In fields language this function interpolates between a set of colors at locations x. While you can be creative about these colors just using another color scale as the basis is easy. For example

designer.color( 256, rainbow(4), x= c( 0,.2,.8,1.0))

leaves the choice of the colors to Dr. R after a thunderstorm. See also colorBrewer to choose sequences of colors that form a good palette.

color.scale assigns colors to a numerical vector in the same way as the image function. This is useful to kept the assigment of colors consistent across several vectors by specifiying a common zlim range.

plotColorScale A simple function to plot a vector of colors to examine their values.

Value

A vector giving the colors in a hexadecimal format, two extra hex digits are added for the alpha channel.

Or in the case of Mines.colors a set of character strings loaded as separate objects with th color name. Use ls() to list them out.

Examples


tim.colors(10) 
# returns an array of 10 character strings encoding colors in hex format

# e.g. (red, green,  blue) values of   (16,255, 239)
#   translates to "#10FFEF" 
# rgb( 16/255, 255/255, 239/255, alpha=.5)
#   gives   "#10FFEF80"  note extra "alpha channel"

# view some color table choices
set.panel( 4,1)
fieldsPlotColors( tim.colors())
title("tim.colors")
fieldsPlotColors( larry.colors())
title("larry.colors")
fieldsPlotColors( two.colors())
title("two.colors")
fieldsPlotColors( snow.colors())
title("snow.colors")

# a bubble plot with some transparency for overlapping dots
set.seed(123)
loc<- matrix( rnorm( 200), 100,2)
Z<- loc[,1] + loc[,2]
colorMap<- color.scale( Z, col=tim.colors(10, alpha=.8))
par( mar=c(5,5,5,5)) # extra room on right for color bar
plot( loc, col=colorMap, pch=16, cex=2)
#  add a color scale
 image.plot(legend.only=TRUE, zlim=range( Z), col=tim.colors(10))

# using tranparency without alpha the image plot would cover points

obj<- list( x= 1:8, y=1:10, z= outer( 1:8, 1:10, "+") )
plot( 1:10,1:10)

image(obj, col=two.colors(alpha=.5), add=TRUE)

coltab<- designer.colors(col=c("blue", "grey", "green"),
                   x= c( 0,.3,1) )
       

image( obj, col= coltab )

# peg colors at some desired quantiles  of data.
# NOTE need 0 and 1 for the color scale to make sense
x<- quantile( c(obj$z), c(0,.25,.5,.75,1.0) )
# scale these to [0,1]
zr<- range( c(obj$z))
x<- (x-zr[1])/ (zr[2] - zr[1])  

coltab<- designer.colors(256,rainbow(5), x)
image( obj$z, col= coltab )
# see image.plot for adding all kinds of legends


data(Mines.colors)
plot( 1:10, 101:110, col=blasterBlue, pch=19)

set.panel()