Histogram Matching demo (cont. #178) (#185)

Bring the scikit histogram matching example into JuliaImages

Co-authored-by: Ashwani Rathee <[email protected]>

Author: johnnychen94

Project.toml

@@ -29,6 +29,7 @@ MosaicViews = "e94cdb99-869f-56ef-bcf0-1ae2bcbe0389"
 Noise = "81d43f40-5267-43b7-ae1c-8b967f377efa"
 OffsetArrays = "6fe1bfb0-de20-5000-8ca7-80f57d26f881"
 PaddedViews = "5432bcbf-9aad-5242-b902-cca2824c8663"
+Plots = "91a5bcdd-55d7-5caf-9e0b-520d859cae80"
 Rotations = "6038ab10-8711-5258-84ad-4b1120ba62dc"
 SimpleTraits = "699a6c99-e7fa-54fc-8d76-47d257e15c1d"
 TestImages = "5e47fb64-e119-507b-a336-dd2b206d9990"
@@ -65,6 +66,7 @@ MosaicViews = "0"
 Noise = "0.2"
 OffsetArrays = "1.0"
 PaddedViews = "0"
+Plots = "1"
 Rotations = "1"
 SimpleTraits = "0"
 TestImages = "0, 1.0"

docs/examples/spatial_transformation/histogram_matching.jl

@@ -0,0 +1,67 @@
+# ---
+# cover: assets/chelsea.gif
+# title: Histogram Matching
+# author: Ashwani Rathee
+# date: 2021-03-23
+# ---
+
+# This demo demonstrates the feature of histogram matching.
+
+# Purpose of Histogram matching is to transform the intensities in a source image so that the
+# intensities distribute according to the histogram of a specified target image. 
+
+# Here, histogram matching is shown for the RGB image. But, it can also be applied on 
+# grayscale images too. Histogram matching can be used to normalize two images when the images 
+# were acquired at the same local illumination (such as shadows) over the same location,
+# but by different sensors, atmospheric conditions, or global illumination. 
+
+using ImageContrastAdjustment
+using Images
+using Plots
+using TestImages
+
+# Load example images: one source image and one reference image 
+
+# !!! note
+#     You need to use `julia >= v"1.3.0"` and `TestImages >= v"1.3.1"` in order to load these
+#     two test images.
+img_source = testimage("chelsea");
+img_reference = testimage("coffee");
+
+# Applying histogram matching on source image using the reference image.
+# Here we use `adjust_histogram` function from ImageContrastAdjustment.jl.
+# It returns a histogram matched image with a granularity of `nbins`, i.e., number of bins. 
+# The first argument `img` is the image to be matched, and the second argument `targetimg` is 
+# the image with the desired histogram to be matched to.
+
+img_transformed = adjust_histogram(img_source, Matching(targetimg = img_reference))
+mosaicview(img_source, img_reference, img_transformed; nrow = 1)
+
+save("assets/chelsea.gif", cat(img_source, img_transformed; dims = 3); fps=2) #src
+
+# To show the effect of histogram matching, we plot for each RGB channel the histogram.
+
+hist_final = [histogram(vec(c.(img)))
+    for c in (red, green, blue)
+    for img in [img_source, img_reference, img_transformed]
+]
+
+plot(
+    hist_final...,
+    layout = (3, 3),
+    size = (800, 800),
+    legend = false,
+    title = ["Source" "Reference" "Histograms Matched"],
+    reuse = false,
+)
+
+# From top to bottom are histograms for the red, green and blue channels.
+# From left to right are the source, reference and the matched images.
+
+# Visual inspection from the plots confirms that `img_transformed` resembles `img_target`
+# much more closely than `img_source`.
+
+# ### Credit and license
+
+# This demo follows the [scikit-image version](https://scikit-image.org/docs/stable/auto_examples/color_exposure/plot_histogram_matching.html#sphx-glr-auto-examples-color-exposure-plot-histogram-matching-py), any usage of
+# this demo should also satisfies the [scikit-image license](https://github.com/scikit-image/scikit-image/blob/main/LICENSE.txt).