Add docs for histogram (boostorg#503)

* Add docs for histogram

* Add docs and make changes:

 - Made changes suggested in first review

 - Add docs for relevant files

* Rename docs file and correct typos

* Change doc for cumulative histogram

* Add docs for histogram equalization

* Make changes suggested in review

* Add docs

* Remove docs for algorithms

* Move images to test_images

Co-authored-by: Mateusz Łoskot <[email protected]>

doc/histogram/create.rst

@@ -0,0 +1,42 @@
+.. _create_histogram:
+
+Create a histogram
+==================
+
+**Method 1** - Using the histogram constructor
+
+Syntax::
+
+    histogram<Type1, Type2, Type3,... TypeN> 
+
+``Type1`` .. ``TypeN`` correspond to the axis type of the N axes in the histogram
+
+Example: If we want a 3D histogram of Axis1 of type ``int``, Axis2 of type ``float`` and Axis3 of type ``std::string``
+we would do it this way::
+
+    histogram<int, float, std::string> h;
+
+And done.
+
+
+**Method 2** (TODO) - Using make_histogram()
+
+There is an alternative to create the histogram directly from
+a GIL image view. 
+
+This should be the preferred over method-1 when creating
+histogram with GIL images, since it creates a histogram with axes configured 
+to match the GIL image. 
+
+Also it is easier than method-1.
+
+Syntax::
+
+    auto h = make_histogram(view(image));
+
+where ``image`` could be a ``gray8_image_t``/``rgb8_image_t`` object read from source.
+
+
+
+
+

doc/histogram/cumulative.rst

@@ -0,0 +1,29 @@
+.. _cumulative_histogram:
+
+Making a cumulative histogram
+=============================
+
+Overview
+--------
+
+A cumulative histogram is a histogram in which each bin stores the count / frequency of itself
+as well as all the bins with keys 'smaller' than the particular bin.
+As such, a notion of ordering among its keys should be existant in the histogram.
+
+The GIL histogram class has the ability to convert itself into its cumulative version. 
+
+Since the container needs to first get an ordering
+over the keys a key sorting takes place before calculating the cumulative histogram. 
+
+Example:
+
+    .. code-block:: cpp
+
+        histogram<int, float> h;
+        /*
+        Fill histogram ...
+        */
+        auto h1 = cumulative_histogram(h);
+
+Tip: *In case you need to store the cumulative histogram elsewhere, consider creating a copy of the histogram
+and then call the function*.

doc/histogram/extend.rst

@@ -0,0 +1,68 @@
+.. _extend_support:
+
+Extending the class
+===================
+
+.. contents::
+   :local:
+   :depth: 1
+
+User defined Axes
+-----------------
+
+In case you need a histogram with an axes of an arbitrary type that is not identified by
+the C++ Standard Library, you need to provide a overload for the hashing function that is 
+used in the histogram class.
+
+GIL's histogram class uses ``boost::hash_combine`` in a sub routine to generate a hash from
+the key.
+
+So we need to provide an overload of ``boost::hash_combine`` for the purpose.
+
+For example, let's consider you need a histogram with an axis over class Test.
+
+.. code-block:: cpp
+
+    // File : ./test.hpp
+    #include <cstddef>
+    #include <functional>
+
+    struct Test 
+    {
+        int a{0};
+        Test() = default;
+        Test(int c) : a(c) {}
+        bool operator==(Test const& other) const
+        {
+            return (a == other.a);
+        }
+    };
+
+    namespace boost {
+        std::size_t hash_value(Test const& t)
+        {
+            // Replace with your hashing code
+            std::hash<int> hasher;
+            return hasher(t.a);
+        }
+    }
+
+Now lets get to the usage example.
+
+.. code-block:: cpp 
+
+    #include <test.hpp> 
+    #include <boost/gil.hpp>
+    #include <iostream>
+    // Mind the order of include i.e. test.hpp before boost/gil.hpp
+
+    using namespace boost::gil;
+
+    int main() 
+    {
+        boost::gil::histogram<Test> h;
+        Test t(1);
+        h(t) = 1;
+        std::cout<<h(t);
+    }
+

doc/histogram/extension/index.rst

@@ -0,0 +1,12 @@
+Extensions
+==========
+
+The GIL documentation sections listed below are dedicated to describe the
+usage of external containers as histograms for GIL images.
+
+.. toctree::
+   :maxdepth: 1
+   :caption: Table of Contents
+
+   overview
+   std

doc/histogram/extension/overview.rst

@@ -0,0 +1,32 @@
+Overview
+========
+
+.. contents::
+   :local:
+   :depth: 1
+
+Description
+-----------
+Apart from the default class supplied by Boost.GIL, there are also provisions to
+use external containers like from std::vector, std::map, boost::histogram etc. These
+are provided as extensions. 
+
+
+Extensions
+----------
+Currently the following are available:
+    #. std::vector (1D histogram support)
+    #. std::map (1D histogram support)
+    #. std::array (1D histogram support)
+    #. std::unordered_map (1D histogram support)
+    #. boost::histogram
+
+
+Adding an external container
+----------------------------
+The workflow should be:
+    #. Provide overloads for fill_histogram(must), make_histogram(optional) etc. in a new file preferably named after the container type in extensions/histogram/.
+    #. Add tests to test/extensions/histogram.
+    #. Add docs to docs/histogram/extensions.
+    #. Other cmake, Jamfile, config etc. file changes.
+

doc/histogram/extension/std.rst

@@ -0,0 +1,45 @@
+.. _std:
+
+STD extension
+=============
+
+Supported Types:
+----------------
+
+#. std\:\:vector (1D)
+#. std\:\:map (1D)
+#. std\:\:unordered_map (1D)
+#. std\:\:array (1D)
+
+
+Usage
+-----
+
+#. **fill_histogram()**
+
+    .. code-block:: cpp
+    
+        // Demo for std::vector
+        std::vector<int> v;
+        gil::gray8_image_t img;
+        /*
+        Fill image ...
+        */
+        gil::fill_histogram(view(img), v, false);
+
+#. **cumulative_histogram()**
+
+    .. code-block:: cpp
+
+        // Demo for std::vector
+        std::vector<int> v;
+        /*
+        Fill vector...
+        */
+        gil::cumulative_histogram(v);
+
+
+
+
+
+

doc/histogram/fill.rst

@@ -0,0 +1,103 @@
+.. _fill_it:
+
+Fill histogram
+==============
+
+.. contents::
+   :local:
+   :depth: 1
+
+Overview
+--------
+
+We will demonstrate the available options for filling an instance of the `histogram` class with
+values that cater from the most simplest to the complex needs that might arise.
+
+Basic
+-----
+
+#. Use operator()
+
+    **Task** - Add value to a particular cell / key / bin in histogram
+
+    .. code-block:: cpp
+
+        histogram<int, int> h;
+        h(1, 2) = 1;
+
+#. Use operator[]
+
+    This requires to input the indices in a format the histogram internally stores its keys,
+    which is of ``std::tuple`` due to its simple interface.
+
+    **Task** - Output value of a bin
+
+    .. code-block:: cpp
+
+        histogram<int, int> h;
+        h(1, 2) = 1;
+        h[{1, 2}] += 1; // Note the curly braces reqd. to construct a tuple
+        std::cout<<h[{1, 2}]; // Output - 2
+    
+#. Use another histogram (copy constructor or assignment)
+
+    **Task** - Fill a histogram using another
+
+    .. code-block:: cpp
+
+        histogram<int, int> A;
+        /*
+        Fill value in A
+        */
+        histogram<int, int> B(A), C;
+        C = A;
+
+#. Use a GIL image view
+
+    You can also use GIL images to directly fill histograms.
+
+    **Task** - Fill histogram using GIL image view
+
+    .. code-block:: cpp
+
+        gil::gray8_image_t img;
+        /*
+        Fill img ...
+        */
+        histogram<int> h;
+        h.fill(view(img));
+        // OR
+        gil::fill_histogram(view(img), h, false); // false if histogram needs to be cleared before filling
+
+
+Advanced
+--------
+
+#. Fill histogram using only a few dimensions of image
+
+    **Task** - Make an histogram over Red and Blue channel of an rgb image
+
+    .. code-block:: cpp
+
+        gil::rgb8_image_t img;
+        /*
+        Fill img ...
+        */
+        histogram<int, int> h;
+        fill_histogram<0, 2>(view(img), h, false); // 0 - red, 1 - green, 2 - blue
+
+#. Fill histogram using GIL pixel
+
+    **Task** - Fill histogram bin using pixel construct in GIL
+
+    .. code-block:: cpp
+
+        gil::gray8_image_t img;
+        /*
+        Fill img ...
+        */
+        histogram<int> h;
+        gil::for_each_pixel(view(img), [](gil::gray8_pixel_t const& p){
+            ++h[h.key_from_pixel(p)];
+        });
+

doc/histogram/index.rst

@@ -0,0 +1,20 @@
+Histogram
+=========
+
+The GIL documentation sections listed below are dedicated to describe the
+histogram class and functions used in many image processing algorithms.
+
+.. toctree::
+   :maxdepth: 1
+   :caption: Table of Contents
+
+   overview
+   create
+   fill
+   subhistogram
+   cumulative
+   stl_compatibility
+   utilities
+   extend
+   limitations
+   extension/index

doc/histogram/limitations.rst

@@ -0,0 +1,6 @@
+.. _limitations:
+
+Limitations
+===========
+
+*TODO*

doc/histogram/overview.rst

@@ -0,0 +1,28 @@
+Overview
+========
+
+.. contents::
+   :local:
+   :depth: 1
+
+Description
+-----------
+
+The histogram class is built on top of std::unordered_map to keep it compatible with other
+STL algorithms. It can support any number of axes (known at compile time i.e. during class
+instantiation). Suitable conversion routines from GIL image constructs to the histogram bin 
+key are shipped with the class itself.
+
+
+Tutorials
+---------
+The following flow is recommended:
+    #. :ref:`create_histogram`
+    #. :ref:`fill_it`
+    #. :ref:`sub_histogram`
+    #. :ref:`cumulative_histogram`
+    #. :ref:`stl_compatibility`
+    #. :ref:`extend_support`
+    #. :ref:`limitations`
+
+.. note:: To try out these tutorials you need to get a clone of the repository, since it is not yet released.