colorResiduals {recolorize}

Calculate squared residuals for color centers

Description

Calculates the squared distance between each pixel and its assigned color center.

Usage

colorResiduals(
  pixel_matrix,
  pixel_assignments,
  centers,
  color_space = "Lab",
  metric = "euclidean",
  ref_white = "D65"
)

Arguments

pixel_matrix

2D matrix of pixels to classify (rows = pixels, columns = channels).

pixel_assignments

A vector of color center assignments for each pixel. Must match the order of pixel_matrix.

centers

A matrix of color centers, with rows as centers and columns as color channels. Rows are assumed to match the index values of pixel_assignments, e.g. a pixel assigned 1 in the assignment vector is assigned to the color in the first row of centers.

color_space

Color space in which to calculate distances. One of "sRGB", "Lab", "Luv", or "XYZ". Passed to convertColor.

metric

Distance metric to be used for calculating pairwise pixel distances in the given color space; passed to dist.

ref_white

Passed to convertColor if ⁠color_space = "Lab⁠. Reference white for CIE Lab space.

Value

A list with the following attributes:

1. sq_residuals: The squared residual for every pixel in pixel_matrix.

2. tot_residuals: The sum of all squared residuals.

3. avg_residual: The average squared residual.

4. residuals_by_center: A list of squared residuals for every color center.

5. avg_by_center: The average squared residual for every color center.

Examples

# RGB extremes (white, black, red, green, blue, yellow, magenta, cyan)
ctrs <- matrix(c(1, 1, 1,
                 0, 0, 0,
                 1, 0, 0,
                 0, 1, 0,
                 0, 0, 1,
                 1, 1, 0,
                 1, 0, 1,
                 0, 1, 1), byrow = TRUE, ncol = 3)

# plot it
recolorize::plotColorPalette(ctrs)

# create a pixel matrix of random colors
pixel_matrix <- matrix(runif(3000), ncol = 3)

# assign pixels
# see `assignPixels` function for details
reassigned <- assignPixels(ctrs, pixel_matrix, adjust_centers = TRUE)

# find residuals from original color centers
color_residuals <- colorResiduals(pixel_matrix = pixel_matrix,
                                  pixel_assignments = reassigned$pixel_assignments,
                                  centers = ctrs)

# compare to residuals from adjusted color centers
color_residuals_adjust <- colorResiduals(pixel_matrix = pixel_matrix,
                                  pixel_assignments = reassigned$pixel_assignments,
                                  centers = reassigned$centers)
# to reset graphical parameters:
current_par <- graphics::par(no.readonly = TRUE)

layout(matrix(1:2, nrow = 2))
hist(color_residuals$sq_residuals,
 breaks = 30, border = NA, col = "tomato",
 xlim = c(0, 1), xlab = "Squared residual",
 main = "Original centers")

hist(color_residuals_adjust$sq_residuals,
breaks = 30, border = NA, col = "cornflowerblue",
xlim = c(0, 1), xlab = "Squared residual",
main = "Adjusted centers")

graphics::par(current_par)