matplotlib · jklymak · Oct 15, 2020 · Oct 15, 2020 · Oct 16, 2020 · Feb 2, 2021
diff --git a/examples/images_contours_and_fields/image_antialiasing.py b/examples/images_contours_and_fields/image_antialiasing.py
@@ -18,6 +18,8 @@
 
 import numpy as np
 import matplotlib.pyplot as plt
+import matplotlib.colors as mcolors
+import matplotlib.cm as cm
 
 ###############################################################################
 # First we generate a 500x500 px image with varying frequency content:
@@ -72,6 +74,81 @@
     ax.set_title(f"interpolation='{interp}'")
 plt.show()
 
+###############################################################################
+# Data antialiasing and RGBA antialiasing
+# ----------------------------------------
+#
+# The examples above all used grayscale.  When colormapping is added there
+# is the complication that downsampling and the antialiasing filters are
+# applied to the data in Matplotlib, before the data is mapped to
+# colors.  So in the following note how the corners fade to white, the middle
+# of the colormap, because the data is being smoothed and the high and low
+# values are averaging out to the middle of the colormap.
+
+f0 = 10
+k = 150
+a = np.sin(np.pi * 2 * (f0 * R + k * R**2 / 2))
+
+fig, axs = plt.subplots(1, 2, figsize=(7, 4), sharex=True, sharey=True,
+                        constrained_layout=True)
+for ax, interp in zip(axs, ['nearest', 'antialiased']):
+    pc = ax.imshow(a, interpolation=interp, cmap='RdBu_r', vmin=-1, vmax=1)
+    ax.set_title(f"'{interp}'")
+fig.colorbar(pc, ax=axs)
+plt.show()
+
+###############################################################################
+# Sometimes, however, we want the antialiasing to occur in RGBA space.  In
+# this case purple is actually the perceptual mixture of red and blue.
+# Matplotlib doesn't allow this to be directly achieved, but we can pass
+# RGBA data to `~.Axes.imshow`, and then the antialiasing filter will be
+# applied to the RGBA data:
+
+fig, axs = plt.subplots(1, 2, figsize=(7, 4), sharex=True, sharey=True,
+                        constrained_layout=True)
+norm = mcolors.Normalize(vmin=-1, vmax=1)
+cmap = cm.RdBu_r
+a_rgba = cmap(norm(a))
+for ax, interp in zip(axs, ['nearest', 'antialiased']):
+    pc = ax.imshow(a_rgba, interpolation=interp)
+    ax.set_title(f"'{interp}'")
+plt.show()
+
+###############################################################################
+# A concrete example of where antialiasing in data space may not be desirable
+# is given here.  The middle circle is all -1 (maps to blue), and the outer
+# large circle is all +1 (maps to red). Data anti-aliasing smooths the
+# large jumps from -1 to +1 and makes some zeros in between that map to white.
+# While this is an accurate smoothing of the data, it is not a perceptually
+# correct antialiasing of the border between red and blue.  The RGBA
+# anti-aliasing smooths in colorspace instead, and creates some purple pixels
+# on the border between the two circles.  While purple is not in the colormap,
+# it indeed makes the transition between the two circles look correct.
+# The same can be argued for the striped region, where the background fades to
+# purple rather than fading to white.
+
+fig, axs = plt.subplots(1, 3, figsize=(5.5, 2.3), sharex=True, sharey=True,
+                        constrained_layout=True)
+f0 = 10
+k = 100
+a = np.sin(np.pi * 2 * (f0 * R + k * R**2 / 2))
+
+aa = a
+aa[np.sqrt(R) < 0.6] = 1
+aa[np.sqrt(R) < 0.5] = -1
+
+norm = mcolors.Normalize(vmin=-1, vmax=1)
+cmap = cm.RdBu_r
+a_rgba = cmap(norm(aa))
+
+axs[0].imshow(aa, interpolation='nearest', cmap='RdBu_r')
+axs[0].set_title('No antialiasing')
+axs[1].imshow(aa, interpolation=interp, cmap='RdBu_r')
+axs[1].set_title('Data antialiasing')
+pc = axs[2].imshow(a_rgba, interpolation=interp)
+axs[2].set_title('RGBA antialiasing')
+plt.show()
+
 
 #############################################################################
 #