ossia::median_filter< T, N > Struct Template Reference

Running median over the last N samples. More...

Detailed Description

template<typename T = float, std::size_t N = 5>
struct ossia::median_filter< T, N >

Running median over the last N samples.

Unlike an average, this rejects isolated outliers outright, which is what one wants against the occasional bogus reading of a sensor or a tracker.

Kept as a chronological ring plus a sorted copy of the same window: an insert is a binary search and a small contiguous move, which for the window sizes this is meant for beats any heap-based scheme by a wide margin.

Note

Feeding a NaN breaks the ordering invariant of the sorted window, and with it every subsequent output. Filter them out upstream.

Public Types
using	value_type = T

Public Member Functions
OSSIA_INLINE void	reset () noexcept
	Forget the history: the next sample restarts the filter.
OSSIA_INLINE void	assign_parameters (const median_filter &) noexcept
	No runtime parameters: the window size is a template argument.
OSSIA_INLINE T	operator() (T x) noexcept

Public Attributes
std::array< T, N >	history {}
	Chronological ring of the window contents.
std::array< T, N >	sorted {}
	The same contents, kept sorted over the first count entries.
std::uint32_t	head {}
	Where the next sample goes in history.
std::uint32_t	count {}
	How many entries are valid, saturating at N.