Image Morphology 2: Convexhull, Filling,Thinning, Clearborder (#203)

Author: ashwanirathee

docs/examples/image_morphology/morphological_operations.jl

@@ -0,0 +1,79 @@
+# ---
+# cover: assets/convex_hull.gif
+# title: Convex Hull, Filling, Thinning, ClearBorder
+# author: Ashwani Rathee
+# date: 2021-7-4
+# ---
+
+# In this demonstration, we will be covering morphological operations like
+# convexhull, image thinning, hole filling and border clearing.
+
+using Images, Noise, ImageMorphology
+using ImageBinarization
+using ImageDraw
+using TestImages
+
+img_src = testimage("morphology_test_512")
+
+# Thresholding of image
+
+img_input = binarize(Gray.(img_src), UnimodalRosin()) .> 0.5
+
+# This is the image that we will be using throughout this demonstration
+# in its boolean form as below mentioned operations are applied on binary images.
+
+Gray.(img_input)
+
+# ## Convex Hull
+
+# Convex Hull operation outputs outer most cover-like boundary of a binary image and
+# returns the vertices of convex hull as a CartesianIndex array.
+
+cordinates = convexhull(img_input)
+img_convex = RGB.(copy(img_input))
+push!(cordinates, cordinates[1])
+draw!(img_convex, Path(cordinates), RGB(1))
+img_convex
+
+# ## Image Filling
+
+# Image filling operation finds connected components of an image using `flood-fill
+# algorithm` and operation gives result image after filling objects that falls in the
+# range of interval specified.
+
+# For filling objects, represent the holes(part to be filled) with `true` in your array. Meaning color RGB(1)/black
+
+# `imfill(img, interval, value, connectivity)` require a `boolean` array, `interval` is a tuple `(a,b)` that specifies the
+# range of values to be filled, the objects of size in this range will be filled with `false` . `connectivity`
+# takes the same values as in label_components (Default value is `1:ndims(img)``)
+
+img_noise = salt_pepper(img_input, 0.5)
+fill_image_1 = imfill(img_noise, (0.1, 1))
+fill_image_2 = imfill(img_noise, (0.1, 10)) # this configuration gets us best results
+fill_image_3 = imfill(img_noise, (1, 10))
+fill_image_4 = imfill(img_noise, (5, 20)) # objects of smaller sizes gets left out
+Gray.([img_noise fill_image_1 fill_image_2 fill_image_3 fill_image_4])
+
+# ## Image thinning
+
+# Thinning operation applies a binary blob thinning operation to achieve a skeletization of the input image.
+# Guo Algorithm, decides which pixels to keep and which to remove using 3 rules given in original paper.
+
+img_thinning = thinning(img_input, algo = GuoAlgo());
+Gray.([img_input img_thinning])
+
+# ## Clear Border
+
+# Clearborder method can be used to clear objects connected to the border of the image.
+
+# `clearborder(img, width, background)` can be used upon binary/grayscale input image
+# on a `width` pixel wide border(`Default is 1 pixel`). `background` is the value that is given
+# to pixels that are cleared(`Default is 0, black color`)
+
+cleared_img_1 = clearborder(img_input, 20); # 20 is the width of border that's examined
+cleared_img_2 = clearborder(img_input, 30); # notice how it remove the inner circle even if it's outside its range
+cleared_img_3 = clearborder(img_input, 30, 1);
+# Default color for removal is 0 meaning remove `RGB(0)` but now since it's 1 it's clears the whole image due to flood fill algorithm
+Gray.([img_input cleared_img_1 cleared_img_2 cleared_img_3])
+
+save("assets/convex_hull.gif", cat(img_input, cleared_img_1, cleared_img_2, cleared_img_3; dims=3); fps=2) #src