Transform

Include <mln/core/algorithm/transform.hpp>

1. void transform(InputImage in, OutputImage out, std::UnaryFunction f)
   

2. image_ch_value_t<InputImage, R> transform(InputImage in, std::UnaryFunction f)
   

   1. Applies the function f to every value of in and stores the result in out.

   2. Same as (1), but out is deduced from input with R = std::decay_t<std::invoke_result_t<UnaryFunction, image_reference_t<InputImage>>>.

   This is equivalent to the following code:

   for (auto&& [vin, vout] : ranges::zip(in.values(), out.values())
       vout = f(vin);
   

   This function has a parallel implementation, see following section for an example.

   Parameters:

   • in – The input image.

   • out – The output image.

   • f – The function to apply on values.

   Template Parameters:

   • InputImage – A model of InputImage

   • OutputImage – A model of OutputImage

   Preconditions

   • in.domain() == out.domain() (1)

   Note

   This function is intended for out-of-place computation. For in-place transformation, see mln::for_each().

Examples

1. Add one to the values of an image:

   mln::image2d<uint8_t> ima = { {1, 2, 3}, {4, 5, 6} };
   auto out = mln::transform(ima, [](uint8_t x) -> uint8_t { return x + 1; });
   

2. Get the first channel of a bi-band image:

   using V = std::pair<uint8_t, uint8_t>;
   mln::image2d<V> ima = {{{1, 0}, {2, 0}, {3, 0}}, {{4, 0}, {5, 0}, {6, 0}}};
   
   mln::image2d<uint8_t> out = mln::transform(ima, &V::first);
   

3. Using parallel transform to add one to the values of an image:

   mln::image2d<uint8_t> ima = { {1, 2, 3}, {4, 5, 6} };
   auto out = mln::parallel::transform(ima, [](uint8_t x) -> uint8_t { return x + 1; });
   

Complexity

Linear in the number of pixels.