gil/test/core/image_processing/simple_kernels.cpp

//
// Copyright 2019 Olzhas Zhumabek <anonymous.from.applecity@gmail.com>
//
// Use, modification and distribution are subject to the Boost Software License,
// Version 1.0. (See accompanying file LICENSE_1_0.txt or copy at
// http://www.boost.org/LICENSE_1_0.txt)
//
#include <boost/gil/image_processing/numeric.hpp>
#include <boost/gil/image.hpp>
#include <boost/gil/image_view.hpp>
#include <boost/gil/typedefs.hpp>
#include <boost/core/lightweight_test.hpp>

namespace gil = boost::gil;

void test_normalized_mean_generation()
{
    gil::gray32f_image_t kernel_image(gil::point_t(5, 5));
    auto view = gil::view(kernel_image);
    gil::generate_normalized_mean(view);
    bool is_correct = true;
    boost::gil::for_each_pixel(view, [&is_correct](gil::gray32f_pixel_t& pixel)
    {
        const auto chosen_channel = std::integral_constant<int, 0>{};
        const auto expected_value = static_cast<float>(1 / 25.f);
        if (pixel.at(chosen_channel) != expected_value)
            is_correct = false;
    });
    BOOST_TEST(is_correct);
}

void test_normalized_mean_throw()
{
    gil::gray32f_image_t kernel_image(gil::point_t(5, 6));
    auto view = gil::view(kernel_image);
    bool have_thrown = false;
    try
    {
        gil::generate_normalized_mean(view);
    } catch(std::invalid_argument& e)
    {
        have_thrown = true;
    }
    BOOST_TEST(have_thrown);
}

void test_unnormalized_mean_generation()
{
    gil::gray32f_image_t kernel_image(gil::point_t(5, 5));
    auto view = gil::view(kernel_image);
    gil::generate_unnormalized_mean(view);
    bool is_correct = true;
    boost::gil::for_each_pixel(view, [&is_correct](gil::gray32f_pixel_t& pixel)
    {
        const auto chosen_channel = std::integral_constant<int, 0>{};
        const auto expected_value = 1.f;
        if (pixel.at(chosen_channel) != expected_value)
            is_correct = false;
    });
    BOOST_TEST(is_correct);
}

void test_unnormalized_mean_throw() {
    gil::gray32f_image_t kernel_image(gil::point_t(5, 6));
    auto view = gil::view(kernel_image);
    bool have_thrown = false;
    try
    {
        gil::generate_unnormalized_mean(view);
    } catch(std::invalid_argument& e)
    {
        have_thrown = true;
    }
    BOOST_TEST(have_thrown);
}

void test_gaussian_kernel_generation()
{
    gil::gray32f_image_t kernel_image(gil::point_t(7, 7));
    auto view = gil::view(kernel_image);
    gil::generate_gaussian_kernel(view, 0.84089642);
    bool is_correct = true;
    const float expected_values[7][7] =
    {
        {0.00000067f, 0.00002292f, 0.00019117f, 0.00038771f, 0.00019117f, 0.00002292f, 0.00000067f},
        {0.00002292f, 0.00078633f, 0.00655965f, 0.01330373f, 0.00655965f, 0.00078633f, 0.00002292f},
        {0.00019117f, 0.00655965f, 0.05472157f, 0.11098164f, 0.05472157f, 0.00655965f, 0.00019117f},
        {0.00038771f, 0.01330373f, 0.11098164f, 0.25508352f, 0.11098164f, 0.01330373f, 0.00038711f},
        {0.00019117f, 0.00655965f, 0.05472157f, 0.11098164f, 0.05472157f, 0.00655965f, 0.00019117f},
        {0.00002292f, 0.00078633f, 0.00655965f, 0.01330373f, 0.00655965f, 0.00078633f, 0.00002292f},
        {0.00000067f, 0.00002292f, 0.00019117f, 0.00038771f, 0.00019117f, 0.00002292f, 0.00000067f}
    };

    const auto chosen_channel = std::integral_constant<int, 0>{};
    for (gil::gray32f_view_t::coord_t y = 0; y < view.height(); ++y)
    {
        for (gil::gray32f_view_t::coord_t x = 0; x < view.width(); ++x)
        {
            auto output = view(x, y).at(chosen_channel);
            auto expected = expected_values[y][x];
            auto percent_difference = std::ceil(std::abs(expected - output) / expected);
            BOOST_TEST(percent_difference < 5);
        }
    }
}

void test_gaussian_kernel_throw()
{
    gil::gray32f_image_t kernel_image(gil::point_t(5, 6));
    auto view = gil::view(kernel_image);
    bool have_thrown = false;
    try
    {
        gil::generate_gaussian_kernel(view, 0.5);
    } catch(std::invalid_argument& e)
    {
        have_thrown = true;
    }
    BOOST_TEST(have_thrown);
}

int main()
{
    test_normalized_mean_generation();
    test_normalized_mean_throw();
    test_unnormalized_mean_generation();
    test_unnormalized_mean_throw();
    test_gaussian_kernel_generation();
    test_gaussian_kernel_throw();
    return boost::report_errors();
}