spice::convolve namespace

#include <include/spice-lib/convolution.hpp>

template<typename T, size_t Channels, size_t Filter_Channels = Channels>

image<T, Channels> spice::convolve::spatial(image<T, Channels> img, image<T, Filter_Channels> filter)

Convolves img with filter

Keep in mind that straight up spatial convolution is expensive. If the filter is separable, i.e. the kernel matrix has a rank of 1, consider using convolution::separable instead.

This function exists mainly as a generalised implementation to be used in the imlpementations of convolution::separable.

template<typename T, size_t Channels, size_t Filter_Channels = Channels>

image<T, Channels> spice::convolve::separable(image<T, Channels> img, image<T, Filter_Channels> filter_h, image<T, Filter_Channels> filter_v)

Convolves img horizontally with filter_h and vertically with filter_v

filter_h is assumed to be of height 1 and filter_v is assumed to be of width 1.

template<typename T, size_t Channels, size_t Filter_Channels = Channels>

image<T, Channels> spice::convolve::separable(image<T, Channels> img, image<T, Filter_Channels> filter)

Convolves img with filter by separating filter into two vectors Fv and Fh such that filter == mul(Fv, Fh)

This function assumes that the passed kernel is separable, i.e. has rank 1

template<typename T, size_t Channels, size_t Filter_Channels = Channels>

image<T, Channels> spice::convolve::frequency_space(image<T, Channels> img, image<T, Filter_Channels> filter, size_t n_threads = 1)

Convolves img with filter by multiplying the two in frequency space.