Padding

Include <mln/core/extension/padding.hpp>

This file provides functions to pad a rectangular buffer or an image either:

inplace: the padding region is defined by some amount of border pixels
in an output buffer: the region is copied and the missing values are filled with the padding policies described below.

Copy the gray area to an output image and fill the missing values according to some padding policies


Padding with a constant value	Padding by edge mirroring	Padding by edge replication	Padding by wraping

enum e_padding_mode

Enum that defines the padding strategies for regular grids. Example is given for a 1D array [0 0 1 2 3 0 0 0 0], borders = {2, 4}

enumerator PAD_ZERO: pad each dimension with value zero [0 0 1 2 3 0 0 0 0]

enumerator PAD_CONSTANT: pad each dimension with a value (here 9) [9 9 1 2 3 9 9 9 9]

enumerator PAD_WRAP: pad each dimension using periodization [2 3 1 2 3 1 2 3 1]

enumerator PAD_MIRROR: pad each dimension using edge mirroring [2 1 1 2 3 3 2 1 1]

enumerator PAD_REPLICATE: pad each dimension using edge replication [1 1 1 2 3 3 3 3 3]

void pad(Image image, e_padding_mode mode, const int borders[][2], T value = {})

template<class T, std::size_t dim> void pad(const ranges::mdspan<T, dim> &buffer, e_padding_mode mode, const int borders[][2], T value = {})

Pad an image or a buffer inplace with the given borders. The borders parameter is a dim × 2 that contains the left/right border width in each dimension that will be filled accorded to the padding policy.


Before padding	After padding

Parameters:

image – The input ndimage (or ndimensional buffer).
mode – The padding policy
borders – A dim × 2 array containing the border widths in each dimension
value – The value used to fill the array (used if mode is PAD_CONSTANT

Example

#include <mln/core/extension/padding.hpp>


int borders[2][2] = {{2,1}, // 1st (x-) dimension
                     {3,2}} // 2nd (y-) dimension
auto input = ...;
mln::pad(input, mln::PAD_CONSTANT, borders, 69);

void copy_pad(Image input, Image output, e_padding_mode mode, T value = {})

template<class T, int dim> void copy_pad(ranges::mdspan<T, dim> input, ranges::mdspan<T, dim> out, mln::ndbox<dim> roi, e_padding_mode mode, T value = {})

Copy an image or a buffer to an external image and fill the missing regions according to the padding policy.


Before padding	After padding

First version

Parameters:

input – The input ndimage (or ndimensional buffer).
output – The input ndimage (or ndimensional buffer).
mode – The padding policy
value – The value used to fill the array (used if mode is PAD_CONSTANT)

Second version

Parameters:

input – The input ndimensional buffer.
out – The output ndimensional buffer.
roi – The roi of the output buffer
mode – The padding policy
value – The value used to fill the array (used if mode is PAD_CONSTANT)

Example

#include <mln/core/extension/padding.hpp>

mln::box2d roi = {0,0,12,12};
mln::box2d roi = {-2,6,12,12};

mln::image2d<uint8_t> input(iroi);
mln::image2d<uint8_t> out(oroi);

mln::copy_pad(input, out, mln::PAD_CONSTANT, 69);