An R port of CDP's Python script to sort RGB color values using principal components analysis

Color_Sort_PCA.R

##### Recreate Cam Davidson Pilon's Python script to sort RGB color values 
#####   using Principal Components Analysis:

library(scatterplot3d)

## Make a function to plot a series of color values in rectangles:
plot_colors <- function(values) {
  # Note, this function takes an matrix of dimensions (n, 3) of numbers
  #   between 0 and 1 for RGB values.  Plotspace is between 0 and 100 for x and y
  x <- 0:100
  y <- 0:100
  n_rects <- nrow(values)
  
  # I'm going to plot the colors by making a bunch of rectangles in a blank plot, then
  #     coloring those rectangles in whatever order that is fed to the function.
  #     What width should each rect take?  Simply, 100 divided by the number of rects!
  width <- 100 / n_rects
  starts <- seq(0, 100-width, length.out=n_rects)             # Starting (lower left) x-coordinates for each rectangle
  plot(x, y, type='n', xaxt='n', yaxt='n', xlab='', ylab='')  # Make a blank plot with plotspace x and y
  for (i in 1:n_rects) {                                      # Loop through the input color values, plotting rectangles
    polygon(x=c(starts[i], starts[i] + width, starts[i] + width, starts[i]),
            y=c(25, 25, 75, 75), col=rgb(values[i,1], values[i,2], values[i,3]))
  }
}

# Some random data:
n = 55
test_mat <- matrix(runif(n * 3, 0, 255), ncol=3, byrow=TRUE) / 255

# Plot the random RGB colors as a 3d plot
par(mfrow=c(1,1))
scatterplot3d(test_mat[,1], test_mat[,2], test_mat[,3], pch=16,
              color=rgb(test_mat[,1], test_mat[,2], test_mat[,3]))

# Do a PCA
pca <- prcomp(test_mat, cor=TRUE)

# Use the values of the first principal component to sort the values of original data
#   and store that in a new matrix (containing the sorted RGB values)
new_order <- order(pca$x[,1])
new_color_mat <- test_mat[new_order,]

## Plot orig and results:
par(mfrow=c(2,1), mgp=c(0,0,0), mar=c(2, 1, 1, 1))
plot_colors(test_mat)
plot_colors(new_color_mat)