Adding a color legend to an image

Question

I have a matrix that I made an image of using image(matrix). Is there away to add a legend of the colors to my image like I do when adding a legend to plot?

Brittani

This question is too vague to be properly answered. Please add programming languages, platforms, libraries, etc. that you are using – Harold

@Harold r is a programming language! – Altamirano

Answer 1

image in R is a fairly basic plotting function. You might want to look at filled.contour if you want a function that will automatically allocate space for a legend. Or try this:

 library(lattice)
 levelplot(matrix)

Walkover

Answer 2

Or the legend could be provided like this:

     legend(grconvertX(0.5, "device"), grconvertY(1, "device"), 
     c("0",".5","1"), fill = colMap[c(1, 10, 20)], xpd = NA)

where grconvertX() and grconvertY() and xpd makes sure the legend is outside the plotting region. A plausible example would be:

    nsamples <- 20
    mat <- rnorm(nsamples, .5, .15)
    dim(mat) <- c(4, 5)
    colMap <- colorRampPalette(c("red","white","blue" ))(nsamples)
    image(1:4, 1:5, mat, col = colMap, ylab="", xlab="")
    legend(grconvertX(0.5, "device"), grconvertY(1, "device"),
    c("0",".5","1"), fill = colMap[c(1, 10, 20)], xpd = NA)

p.s.: I know it is an old request and it is solved. However I was looking for a similar answer and I could not find it. Since I bother solving this issue I thought maybe someone else could also benefit from it.

Vassal

Very helpful. For a plot with multiple figures, I used par(mfrow=c(1,1), new=TRUE, fig=c(0,1,0,1)) in front of the legend statement and "nfc" rather than "device". – Inflated

When image() interprets the color vector though, it must be mapping from your data min-to-max, so probably need to adjust the legend values to match the data exactly? e.g. 0 and 1 don't necessarily exist in your example data, so the legend may not be mapping the colors exactly. – Phatic

Answer 3

From the package fields, you could try image.plot. This function is based on the regular image, but it provides a figure legend.

library(fields)
x = 1:10
y = 1:15
z = outer( x,y,"+") 
image.plot(x, y, z)

Lagan

Answer 4

To make a gradient color bar, you can do this:

# make example data and color key
x <- matrix(data = sample(x = 5:85, size = 100, replace = T), nrow = 10)
n.colors <- 90
color.fun <- colorRampPalette(colors = c("magenta2", "grey"), bias = 10)
col.key <- data.table("color" = color.fun(n = n.colors),
                      "value" = seq(from = min(x), to = max(x), along.with = 1:n.colors))

# make the plot with image- leaving space for legend using fig
par(fig = c(0,.9,0,1), mar = c(2,2,2,0))
image(z = t(x), axes = F, ann = F, col = col.key$color)

# add the legend
par(fig = c(.9,1,.3,.7), mar = c(1,1,1,2.5), new = T)
plot(x = rep(1,length(col.key$value)), y = col.key$value, xlim = c(0,1), col = col.key$color, type = "n", xaxs = "i", yaxs = "i", ann = F, axes = F)
segments(x0 = 0, x1 = 1, y0 = col.key$value, y1 = col.key$value, col = col.key$color, lwd = 5)
axis(side = 4,lwd = 0, las = 2, line = -.75)
mtext(text = "Legend", adj = 0, line = 1, cex = 1.2)

Note that this also addresses I think an error in HelloWorld's answer, because you want the colors to exactly map between the color bar and the image (so scale must match your actual data values).

Phatic