Transform (View)

Include <mln/core/image/view/transform.hpp>

1. auto transform(Image ima, std::UnaryFunction f)
   

2. auto transform(Image ima, Image ima2, BinaryFunction f)
   

   1. Makes a view from ima where for each pixel value evals to out(p) = f(ima(p))

   2. Makes a view from ima where for each pixel value evals to out(p) = f(ima(p), ima2(p)). For this overload, the domain of both image must be equal.

   Parameters:

   • ima – Input range

   • ima2 – Input range

   • f – Function to apply on pixel values

Warning

The function f must not create dangling references on the input.

mln::image2d<mln::rgb8> ima = ...;
auto g1 = mln::view::transform(ima, [](rgb8 v) { return v; }); // OK: identity but creates a non-mutable image
auto g2 = mln::view::transform(g1, [](const rgb8& v) -> const uint8& { return v[0]; }); // KO: create a dangling reference to g1(p)

		ima, ima2	transform(ima, f)	transform(ima, ima2, f)
Category	Forward	X	X	X
	Bidirectional	X	X	X
	Raw	X
Properties	Writable	X	X*	X*
	Accessible	X	X	X
	Indexable	X	X	X

(*): only if f is a projection.

Examples

• Multiply all values by 2:

mln::image2d<uint8_t> input = { {0, 1, 2}, {3, 4, 5} };
auto g = mln::view::transform(input, [](mln::uint8 v) { return v * 2; });
mln::io::imprint(g);

Outputs:

0 2 4
6 8 10

• Write on the red projected member:

mln::image2d<mln::rgb8> input = {{{0, 0, 0}, {1, 1, 1}, {2, 2, 2}}, {{3, 3, 3}, {4, 4, 4}, {5, 5, 5}}};
auto g = mln::view::transform(input, [](mln::rgb8& x) -> uint8_t& { return x[0]; });
mln::fill(g, 42);
mln::io::imprint(g);

Outputs:

[42,0,0] [42,1,1] [42,2,2]
[42,3,3] [42,4,4] [42,5,5]

• Transform two images with a binary function (alpha blending):

mln::image2d<uint8_t> input1 = ...;
mln::image2d<uint8_t> input2 = ...;
auto f = [](uint8_t x, uint8_t y) { return alpha * x + (1 - alpha) * y; };
auto g = mln::view::transform(input1, input2, f);