What is a color map?

• A color map is a function from input values to colors used for plotting.
• The input values might be discrete, e.g., {1, 2, 3, 4} or continuous, e.g, [0.0 .. 1000.0]
• The colors might be simply grayscale brightness, or an RGB triplet (or HSV, or ...)
• The colors might be used for an image, or for a collection of lines, symbols, etc.

Types of color map

• Sequential maps – for unsigned scalar values, e.g., brightness
• Divergent maps – for signed scalar values, e.g., relative velocity
• Categorical maps – for a collection of unordered items, e.g., different galaxies
• Circular maps – for values that wrap around continuously, e.g., angles

Sequential maps

Most desirable property:

Monotonic mapping between value and color brightness

Why? Because brightness is the aspect of color that we intuitively understand. We want to leverage all the good work that evolution has done in designing our visual system.

Classic example:

simple grayscale

In many cases, grayscale is the best pick.

%matplotlib inline
from matplotlib import pyplot as plt
import seaborn as sns
import numpy as np
from astropy.io import fits
sns.set_context('poster')

hdu = fits.open("data/OrionS-Rubin-WFC3-F658N.fits")[0]
data = hdu.data[0:1000, 0:1000]

def imshow(cmap):
    fig, ax = plt.subplots(figsize=(6, 5))
    im = ax.imshow(data, origin="lower", 
              vmin=0, vmax=20.0, 
              interpolation="none", cmap=cmap)
    fig.colorbar(im, ax=ax)

Example grayscale image

imshow("gray")

Example negative grayscale image

imshow("gray_r")

Note that the positive grayscale accentuates the brighter regions, while the negative grayscale accentuates the fainter regions.

Other desirable properties

Perceptually uniform

Equal changes in input values produce equal changes in color.

More than just brightness varies across the scale

For instance, hue or saturation might change. This makes it easier to pick off a particular value.

Examples of perceptually uniform sequential color maps

For more info, see the Choosing Colormaps in Matplotlib tutorial.

Viridis

We lose a little in contrast, but the qualitative blue–green–yellow variation allows us to better cross-reference with the colorbar to see the approximate values of the individual pixels.

imshow("viridis")

Inferno

I like this one better.

imshow("inferno")

Cividis

This is optimized with consideration for people with red–green color vision deficiency. See paper by Nuñez et al. (2018)

imshow("cividis")

What not to do

Never ever use the jet color map

Not perceptually uniform. Not even monotonic in brightness. There are scientific studies that prove that this color map is bad

imshow("jet")