OpenEV
Extending OpenCV to event-based vision
Loading...
Searching...
No Matches
ev::BackgroundActivityFilter Class Reference

Background activity filter for event noise removal. More...

#include <filtering.hpp>

Public Member Functions

 BackgroundActivityFilter (const cv::Size &size, ev::TimeType dt, int radius=1)
 BackgroundActivityFilter (const BackgroundActivityFilter &)=default
 BackgroundActivityFilter (BackgroundActivityFilter &&) noexcept=default
BackgroundActivityFilteroperator= (const BackgroundActivityFilter &)=default
BackgroundActivityFilteroperator= (BackgroundActivityFilter &&) noexcept=default
bool operator() (const ev::Event &e)

Detailed Description

Background activity filter for event noise removal.

Discards isolated events that have no correlated neighbor within a spatial radius and time window. An event passes if at least one pixel in its neighborhood fired within the last dt time units.

Note
T. Delbruck, "Frame-free dynamic digital vision"

Constructor & Destructor Documentation

◆ BackgroundActivityFilter()

ev::BackgroundActivityFilter::BackgroundActivityFilter ( const cv::Size & size,
ev::TimeType dt,
int radius = 1 )

Constructor.

Parameters
sizeSensor resolution (width × height)
dtTime threshold; events with no neighbor firing within this window are discarded
radiusSpatial neighborhood half-size (default 1 → 3×3 / 8-connected)

Member Function Documentation

◆ operator()()

bool ev::BackgroundActivityFilter::operator() ( const ev::Event & e)
nodiscard

Test and record a single event.

Parameters
eEvent to evaluate
Returns
True if the event passes (at least one neighbor fired within dt)
Note
The internal timestamp map is updated regardless of whether the event passes.

The documentation for this class was generated from the following files: